From qualitative to quantitative interpretation: An interpreter’sguide to fluid prediction in Pliocene to Turonian deepwaterturbidites from West Africa to Asia Pacific

Uwe Strecker1, Paola Vera de Newton1, and Maggie Smith1

Abstract

To mitigate exploration risk in deepwater settings, subsurface analysis increasingly has to rely on integrationof qualitative with quantitative techniques. To predict pay in turbidite sandstones, proven statistical and ana-lytical methods can routinely be run on well and seismic inversion data. However, quantitative interpretation(QI) should begin with a responsible audit of available well logs and seismic data, succeeded by data condition-ing, proceeding with quality control, and placing elastic attribute responses within their geologic context. Toaddress these issues, we evaluate geologic controls on porosity change as manifested by overpressure andcompaction on calibration and analysis of elastic attributes. Following calibration of seismic inversion data,we provide tutorial-style interpretations of deepwater clastic reservoirs from the Gulf of Guinea, West Africa,to the Sabah trough, Borneo. Case study examples offer interpreters the potential to use workflows surroundingdata mining in exploration or during field development. In our first example, a comparison of univariate sta-tistics run on compressional- and shear-wave impedances and Poisson’s ratio is introduced to potentially datamine 3D seismic over turbidite fairways. Joint interpretation of P-wave and S-wave impedances is combinedwith innovative uses of bivariate statistical analysis for anomaly detection. Additionally, the geologic rationaleof interpreting elastic relationships of calibrated attributes, such as Lambda Rho and Mu Rho, is discussed onthe seismic scale of a single reservoir layer using a combination of statistical methods and rock physics. Here,qualitative interpretation, via application of principles from seismic stratigraphy and seismic geomorphology,ultimately unlocks ambiguity in rock-physics-driven, quantitative lithology determination, guiding application ofQI routines toward correctly predicting the prevailing fluid type. Elastic calibration permits seismic lithofaciesclassification of Cretaceous turbidite sandstones deposited as middle to lower slope channels canyon-fill andbasin-floor channel complexes.

IntroductionBecause interpretation of seismic attributes is poten-

tially ambiguous (Chopra and Marfurt, 2008), this paperprovides interpreters with data-reduction techniques tofirst quality-control seismic inversion results. One suchqualitative technique contrasts compressional- and shear-wave impedance histograms to potentially automate 3Dseismic data mining in turbidite fairways (e.g., Barnes,2001). The following two sections on methodologydiscuss geologic effects such as overpressure, compac-tion, and cementation on calibration and analysis ofseismic inversion products. Finally, quantitative inter-pretation (QI) of elastic seismic attributes is discussedon the scale of a seismically resolved bed (e.g., Widess,1973) using a combination of statistical methods androck physics (e.g., Beaubouef et al., 1999). Here, draw-ing upon qualitative stratigraphic principles and seismic

geomorphology eliminates invalid alternatives for lith-ology determination, guiding QI toward the correctprediction of prevailing fluid types.

For practical reasons, numerical models of mostnatural systems (weather forecasts, ocean currents,etc.) can only be approximated by complex functionsof position and/or time (Oldenburg and Ellis, 1991;Parker, 1994). Seismic inversion aims to retrieve sub-surface properties by building an integrated earthmodel that physically explains the recorded acousticsignals of a finite data set, especially reservoir charac-teristics (Parker, 1994). However, due to its complexity,the inversion process may be prone to a series of grosserrors arising from (1) poor data quality; i.e., noise, in-sufficient multiple suppression, frequency distortion ofCMP gathers after NMO correction, etc. (Yilmaz, 2001),(2) suboptimal inversion parameterization; i.e., poorly

1Rock Solid Images, Houston, Texas, USA. E-mail: uwe.strecker@rocksolidimages.com; paola.newton@rocksolidimages.com; maggie.smith@rocksolidimages.com.

Manuscript received by the Editor 12 July 2013; revised manuscript received 10 October 2013; published online 11 February 2014. This paperappears in Interpretation, Vol. 2, No. 1 (February 2014); p. SA127–SA140, 13 FIGS.

http://dx.doi.org/10.1190/INT-2013-0106.1. © 2014 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

t

Special section: Seismic attributes

Interpretation / February 2014 SA127Interpretation / February 2014 SA127

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

calibrated seismic interval velocities, selecting a spa-tially invariant wavelet, or incorrect well locations;(3) ignoring anisotropy effects on low-frequency modeltrends, or (4) an otherwise flawed earth model (i.e., inunconsolidated sediment, seismic velocity changes mayactually more closely follow the water bottom geometryand conform to structure only with increasing compac-tion depth (R. Arora and M. Suda, personal communi-cation, 2011). Such errors must be remedied duringquality control preceding the actual interpretation proc-ess to avoid interpretation on substandard seismiccubes. Many types of seismic inversions are availableto the interpreter today (for example, simulatedannealing [Pedersen, 1990]; elastic inversion [Connolly,1999]; extended elastic inversion [Whitcombe et al.,2002]; colored inversion [Lancaster and Whitcombe,2000]; simultaneous impedance inversion [Tonellotet al., 2001]; Bayesian linearized AVO inversion [Bulandand Omre, 2003], and petrophysical seismic inversion[Coléou et al., 2005]). Some prestack inversion applica-tions produce either seismic impedances and a densityvolume, as deliverables (for example, Tonellot et al.,2001), or, alternatively, a VPV−1

S volume, a density vol-ume and a P-wave or S-wave impedance volume, insteadof directly inverting for reservoir properties, such aswater saturation or VClay (Whitcombe et al., 2002). Thispaper focuses on interpretation of seismic impedancesand their elastic constructs, such as the Lambda Rhoand Mu Rho attributes (Goodway et al., 1997).

The primary objective of this article is to supplysubsurface interpreters with statistical and analyticalworkflows to effectively evaluate seismic signaturesat the reservoir scale derived from cardinal relation-ships that exist between elastic properties. Recentadvances in the characterization of fine-grained turbi-dite systems integrate applied subsurface interpretationparadigms such as sequence stratigraphy and seismic

geomorphology with reservoir analyses, especiallyas related to clastic deepwater settings (Weimer et al.,1998; Beauboeuf et al., 1999; Gardner and Borer, 2000;Posamentier and Kolla, 2003; Figure 1).

MethodsA phased approach to QI of seismic inversion data

is presented that includes a brief discussion of (a) dataquality control, (b) application of univariate statistics toseismic data mining (histogram matching), (c) optimalvisualization of a Class II(P) AVO anomaly in P-waveand S-wave impedances, (d) bivariate statistical analy-sis, (e) detrending elastic data for compaction effects,and (f) QI using rock physics to predict prevailingreservoir fluid type from analysis of elastic seismicattributes.

Quality controlOrdinarily, it is taken for granted that all seismic and

well log data have been properly conditioned beforebeing incorporated into an inversion. In practice, thisrequires that all seismic gathers have been properlyconditioned (Singleton, 2009) for amplitude preserva-tion, correct phase, statics, NMO stretch, residualmove-out, and multiple suppression. In the case of welldata, erroneous or missing wireline curve data intro-duced by bore hole breakouts, washouts, cycle skips,bed corrections, and mud-filtrate invasion must besuccessfully corrected or estimated (Walls et al., 2004).

These precursory steps are instrumental in generat-ing a geologically sound low-frequency earth model.Thus, to assess the interpreter’s fidelity in the qualityof the inversion, data generated during inversion shouldindependently be verified, for instance, by comparingcrossplots of elastic attributes from upscaled log curvesagainst their seismic counterparts. Specifically, an in-terpreter may elect to address the following contingentissues prior to QI:

• Do elastic property values from upscaled welllogs and seismic data fall within similar dynamicranges?

• Do the lithology trends for sand, silt, shale, car-bonate (chalk, dolostone, limestone, marl), vol-canics, chert, etc., of upscaled log curves andinversion attributes match?

• Do log and inversion anomalies form coincidentclusters?

Only if all conditions are met is the inversion consid-ered to be “in gauge,” i.e., within acceptable tolerancesand ideallymatching crossplots of upscaled logs. Figure 2contrasts a crossplot of Lambda Rho versus Mu Rhousing the upscaled well log data (Figure 2a) with thesame elastic attribute pairs calculated from simultane-ous inversion products (Figure 2b). The shale baselinein the upscaled log data delivers a perfect matchwith the shale baseline trend seen in the crossplotfeaturing seismic samples (Figure 2). This observation

Figure 1. Perspective view of Pliocene Isongo turbiditecomplex, offshore West Africa, λρ (“Lambda Rho”) seismicattribute, color scale in GPa � gcm−3. Fluid prediction fromvisualization of a single attribute can prove highly uncertainunless it is ably supported by QI.

SA128 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

is significant because two independently derived back-ground trends — the one derived from the simultane-ous inversion via the integrated earth model (commonlyseismic interval velocity plus well data; Figure 2b) andthe one derived from upscaled well-log calibration [Fig-ure 2a]) — have produced congruent results. Althoughthe inversion generally appears to be in gauge, — i.e.,sands in the seismic data cluster occur at an equal dis-tance from the shale baseline as do sand samples in thecalibration well — most seismically predicted sandsusing the hybrid attribute show somewhat lower attrib-ute value pairs than the equivalent sand samples fromthe upscaled logs of the calibration well. A cluster ofsand samples from the upscaled log curves (i.e., MuRho values from 40 to 50 [GPa � gcm−3]) is not observedat an equivalent position of seismic samples interpretedas sand, i.e., Mu Rho values from 20 to 30 [GPa � gcm−3]and Lambda Rho values from ca. 2–20 [GPa � gcm−3](Figure 2). This subtle dichotomy is most likely causedby a change in pressure regime between the analyzedseismic data and the much shallower calibration welllocated outside the boundaries of the 3D prestacksurvey area, and, therefore, well data appear only mar-ginally suited for a direct comparison. Nevertheless,lower elastic values found at greater depths than thoseencountered in sand-rich intervals of the shallow cali-bration well are attributed to overpressure as evidencedby low resistivity and density values and high intervaltransit times on sonic logs from elsewhere in this Asianback arc region prone to host overpressured compart-ments. Poisson’s ratio values from suchintervals tend to be anomalously highand instead resemble data values fromshallower sections (W. Marin, personalcommunication, 2013). This observationunderscores the vital importance of ac-counting for geologic factors, such asoverpressure, compaction, or cementa-tion on porosity during calibration ofseismic data (e.g., Steckler and Watts,1978; Ehrenberg, 2004; Clavaud, 2008).High Lambda Rho seismic attribute val-ues >45 [GPa � gcm − 3] not encoun-tered in the upscaled log data fromthe calibration well (high on structure)suggest presence of calcareous shaleor, possibly, marls, imaged only by theinversion data from deeper within the3D seismic survey.

Ultimately, however, shale and sandtrends alike will have to match betweenthe upscaled well log crossplot data andthe seismic crossplot representation ofthe elastic data to verify that the inversionhas been successful. To ascertain thatthe calibration of the sand probabilityattribute has been successful, a log suiteconsisting of VClay (black) and upscaledlog curves of Lambda Rho (magenta)

and Mu Rho (the latter attribute shown with an invertedscale to match the VClay deflection toward lower valuesin sands; orange curve) are compared to an insert rib-bon-band display of the sand probability indicator attrib-ute invariable-densitymode(Figure3).All lowVClay zones(sand-rich beds) become successfully highlighted in gold-brown colors derived from the relational crossplot colorfor high-probability sand (highest probability for sandattribute value ¼ 255). In contrast, shale-prone log sec-tions are highlighted by various shades of gray. Forbest results, elastic calibration to sand should be tied toseismic resolution, i.e., tuning analysis (which is outsidethe scope of this paper).

Uses of univariate statisticsin seismic data mining

Univariate statistics provides a useful starting pointfor joint data mining of seismic impedance volumes.Body wave velocities, such as P-wave and S-wave veloc-ities, essentially track one another except in the pres-ence of hydrocarbons and/or fractures (Avseth et al.,2005). Thus, a pattern of disproportionate deviationin either P-wave impedance, S-wave impedance, or bothimpedances potentially bears valuable information forthe interpreter: For example, retardation of S-wavevelocity compared to P-wave velocity could suggestfaulting and/or seismically discernible large-scale frac-turing, whereas a disproportionate excursion of P-waveimpedance toward lower attribute values could indicatethe presence of hydrocarbons (e.g., Batzle and Gardner,

Figure 2. Quality control of seismic inversion data through bivariate analysis.Comparison of upscaled Lambda Rho versus Mu Rho log data from calibrationwell (a) and seismic data (b). Generally, upscaled log and seismic data samplesappear in brownish color, only nonbrownish colors indicate a higher count ofsamples. The variable density background relation featured in each crossplotdenotes the increasing probability of encountering sand normal to a shiftedsand baseline or shale baseline. Matching background trends of upscaledlog curve data and seismic samples for shale base lines in both crossplots(gray-white trend lines) confirm that the inversion is calibrated within accept-able tolerance levels.

Interpretation / February 2014 SA129

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

2000). To expand this concept to data mining, a histo-gram of S-wave impedance is contrasted with one of P-wave impedance from a West African deepwater setting(Figure 4). For easier comparison, the base width ofone of the impedance histogram distributions is ad-justed to approximately match the other impedancehistogram in overall shape. Because body waves travelessentially “in tandem,” their impedance histogram dis-tributions are expected to match over depth compart-ments with approximately constant VP∕VS, i.e., thenonfluid effects on P-wave and S-wave velocities shouldproduce similar histograms (e.g., Avseth et al., 2005).However, this comparison reveals that a series of fourpeaks in the upper histogram is not matched by an equalnumber of peaks present in the lower histogram. In thelower histogram, there are only three peaks, and manyof the seismic data samples characterized by higher

shear impedances have shifted toward lower P-waveimpedances in the lower histogram, effectively infillinga trough that exists at the corresponding location in theS-wave impedance histogram. This dichotomy stronglysuggests sensitivity of the analyzed seismic data to flu-ids, because in the typical case where all of these rocksare wet and unfractured, the relative abundances ofdata impedances should match for both histograms.To further test this hypothesis, additional inspectionof a third histogram of attribute values for the same dataset (third horizontal panel in Figure 4) reveals a nega-tive skew in Poisson’s ratio toward the leading edge ofthe distribution. Because a single population of anyphysically measured property would most likely obeya Gaussian distribution, the observed lognormal skewin the data must have been introduced by yet anothercontrolling factor. Interpreters familiar with coherence(or other multitrace attributes like similarity, variance,maximum curvature, etc.) immediately recognize thatthe distribution is skewed toward where any of theaforementioned multitrace attributes would detectfaults or stratigraphic boundaries that originated fromphase breaks (“edges”) inherent in the seismic data(Bahorich and Farmer, 1995). Hence, the skew presentin Poisson’s ratio must likewise be attributed to avolumetrically small, yet noticeable, anomalous “edge”in the data. Parsimony dictates that this leading edge,negatively skewed histogram signature in Poisson’sratio most likely stems from sensitivity of the prestackseismic data to hydrocarbons, an interpretation benefitgleaned from simultaneous inversion (Figure 4). How-ever, if the hydrocarbon anomaly were large enough,the histogram distribution would return to a Gaussianshape, but this signature may never be observed withinlarge 3D seismic surveys.

Optimizing attribute visualization using rockphysics knowledge

At a minimum, analysis of simultaneous inversiondata should seek to jointly interpret P-wave- and S-waveseismic impedances. Specifically, an interpreter’s aimshould be to adjust the dynamic ranges of each seismicimpedance such as to obtain comparable backgroundcoloration. To color backgrounds evenly serves as a vis-ual filter intended to make it nearly impossible to dis-cern which of the two impedances is on display, wouldit not be for the attribute values identified in the colorbar; the rationale being that only attribute values fromlayers differing in seismic expression between the twoimpedance volumes emerge more readily. After colorbar manipulation, P-wave and S-wave seismic imped-ance sections will appear nearly identical in overallexpression with the exception of anomalies (Figure 5):notably, the P-wave impedance anomalies near the topof the section have disproportionately lower valuesthan background, whereas the corresponding S-waveimpedance section features slightly elevated layerproperties there (e.g., Batzle and Gardner, 2000, theirFigure 15). Both observations are compatible with the

Figure 3. Elastic calibration to sand. Well elastic relationsfor log-scale VClay (black curve) and upscaled Lambda Rho(magenta) and Mu Rho (orange) curves are shown. The MuRho curve scale has been inverted to match the sand deflec-tions of the VClay curve. Interestingly, the Mu Rho attributedoes not unequivocally detect sands, as a silt streak belowthe reservoir sands features comparable elastic values (bluearrow). (b) The display features an insert of the relational hy-brid sand probability attribute that retains the cardinal rela-tionships between Lambda Rho versus Mu Rho exhibited incrossplot space (superposed multicolored ribbon band) (b).All low VClay zones (sand-rich beds) become successfully litin this calibration, as is evidenced by gold-brown colorsderived from the relational crossplot color for high-probabilitysand (highest probability for sand attribute value ¼ 255). Incontrast, shale-prone log sections are highlighted by variousshades of gray.

SA130 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

elastic response of either an oil-bearing reservoir sandor a gas sand because the presence of hydrocarbonslowers reservoir density and retards P-wave propaga-tion (via drastically lowering the fluid bulk modulus,Kf ), but hydrocarbon presence will also increase theS-wave velocity in this zone (Figure 6). This is because

division of the shear modulus by a smaller number thanwater density results in an increase in S-wave velocityover that of the wet case (Figure 6). For that reason, thereduced bulk density in the hydrocarbon case barelylowers the shear impedance compared to its wet state(Figure 6). However, the primary reason why S-wave

Figure 5. Elastic impedances from simultaneous inversion. Both seismic impedance sections resemble one another closely withthe exception of a few deviatory areas. Areas 1 and 2 identify localized anomalies characterized by a deviation toward dispro-portionately lower P-wave impedance values than background, coupled with a slight increase of S-wave impedance values overbackground. Another minor anomaly similar to the previous two is located ca. 100 [ms] below area 2. All anomalies are interpretedas potentially hydrocarbon-bearing (see text for discussion). In contrast, area 3 identifies a high-impedance layer characterized byvery high P-wave impedance, yet low S-wave impedance values. This seismic signature is compatible with the interpretation of thisfeature as a fractured marl, a common lithology for this region in the Gulf of Guinea.

Figure 4. Univariate analysis of P-wave-impedances and S-wave impedances in thecontext of body wave velocities (d). Colorbar denotes voxel density (as does y-axis).Horizontal axis denotes P-wave impedancein (a), S-wave impedance in (b), and Poisson’sratio in (c). Conspicuously, a low abundanceof S-wave impedance values at 2250 [mgcm−3](a) is coupled with a much higher abundanceof low P-wave impedances in the histogrambelow (b) suggesting sensitivity of the seismicdata to fluids because a gas-charged sand willrapidly drop its compressional velocity whilesimultaneously experiencing a slight increasein S-velocity (a). A histogram of seismic sam-ples from the resultant Poisson’s ratio volumelikewise exhibits an “edge effect” believed toresult from hydrocarbon-charged reservoirrock evoking a “log-normal” skew at the lead-ing edge of the distribution.

Interpretation / February 2014 SA131

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

impedance is higher for the wet case and the hydrocar-bon case, is because the shear modulus of quartz ishigher on average than the shear moduli of the shale.As a consequence, a sand-prone reservoir section po-tentially increases in S-wave impedance over a shalesection poorer in quartz content. However, becausethe numerical change in bulk density is small comparedto the large reduction in P-wave velocity, the densitydeviation is only very subtle for either of the fluid-sub-stituted S-wave impedance log curves (Figure 6). Thelogplot example shown is a classic example of anAVO Class II(P) anomaly that would be difficult to spoton poststack amplitude data (Figure 6). In summary:whereas the observed seismic anomalies primarily havevalues much lower than background on the P-waveimpedance section (Figure 5), the same interval hasslightly higher than background shear-wave impedancevalues due to lithology changing from shale to sand(Figure 6). One of the key observations from fluid-substitution is that S-wave seismic impedances, unlikeP-wave seismic impedances, can be expected to be vir-tually identical to one another for either wet or gascases (Figure 6). This observation justifies the claimfrom the previous univariate analysis that treats S-waveimpedances for wet or gas-saturated rock as essentially

stationary, whereas P-wave impedances significantlydiffer for each fluid-substituted case. In a qualitativesense, the biconvex seismic morphology of thesetwo mid-Cretaceous anomalies suggests differentialcompaction of (channel) lenses encased in shale(e.g., Chopra and Marfurt, 2012) with the larger ofthe two anomalies being located in a trapping configu-ration on the up-thrown side of a normal fault (Figure 5).A third anomaly of opposite acoustic character (i.e.,high P-wave impedance, low S-wave impedance) isinterpreted as fractured marl, a deepwater lithologycommon to this West African offshore region (Figure 5).

QI using fundamental statisticsHaving thus identified two potentially hydrocarbon-

charged reservoir sand lenses by scanning for colordifferences over background between these two seis-mic impedance attribute sections, we proceed to testhow anomalous these features really are. The visualdiscrimination process outlined (Figure 5) could beformalized by attribute normalization to create a differ-ence volume (untested here). For the purpose of furtherstatistical analysis, we map the entire seismic horizoncontaining the two anomalies instead, and subsequentlyextract a hereditary, i.e., mathematically derived seis-

mic attribute (Russell et al., 2003) fromthese impedance sections along a tra-verse (Figure 7). For this exampleLambda Rho and Mu Rho are calculatedfrom the impedance data for furtheranalysis (e.g., Chopra and Pruden,2003). Lambda Rho is obtained by takingthe difference between the square of theP-wave impedance minus double thesquare of the Shear wave impedance(Russell et al., 2003; Chopra and Pruden,2003; Avseth et al., 2005). The elasticconstants λ (Lambda) and μ (Mu),a.k.a. Lamé’s constants, play an impor-tant role in (extended) elastic inversion(Whitcombe et al., 2002), but are not fur-ther discussed here. Lambda Rho (prod-uct of λ and ρ, bulk density) is referredto as the “incompressibility” attributethat, in addition to porosity changewithin a lithology, potentially owes afraction of its seismic response to hydro-carbon charge (Chopra and Pruden,2003; Li et al., 2003). A baseline repre-senting the mean through the datais plotted on the traverse and threestandard deviations are subsequentlysubtracted from the mean (Figure 7).Seismic attribute values located threenegative standard deviations below themean do not conform to the statisticaldefinition of a Gaussian distributionand consequently represent a lithologi-cal, porosity, and/or fluid anomaly of a

Figure 6. Log plot of in situ gas sand (black) versus wet (blue) fluid-substitutedelastic curves. Track 1: identifies mineral volumes, i.e., VClay; VCalcite; VQuartz;track 2: pairs effective porosity and water saturation; track 3: bulk density, 2to 3 gcm−3; track 4: P-wave velocity, 2,000 to 6; 000 ms−1; track 5: S-wave veloc-ity, 1,000 to 3; 000 ms−1; track 6: P-wave impedance, 6 to 12 kms−1 gcm−3, track7: S-wave impedance, 3 to 6 kms−1 gcm−3; track 8: VP∕V−1

S ; track 9: Poisson’sratio, track 10: Lambda Rho, 10 to 60 [GPa � gcm−3]; track 11: Mu Rho, 10 to40 [GPa � gcm−3]; track 12: NMO-corrected synthetic seismogram offset gather,in situ; track 13: NMO-corrected synthetic seismogram offset gather, wet condi-tion. Of the elastic attributes shown, Lambda Rho offers the most discriminationto predict pay, whereas the S-wave impedance or its squared product with bulkdensity (Mu Rho) offer much warranted lithology discrimination. Both NMO-corrected offset gathers feature a Class II(P) AVO response with the addedcaveat of the wet case stacking to nearly zero amplitude. In contrast, the insitu gas case features a strong trough before reaching critical, facilitating therecognition of this anomaly in poststack amplitude data. The figure givesvaluable insight into seismic resolution.

SA132 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

different distribution (Figure 7), i.e., statistically thecenter of the anomalies contains elastic values thatcannot be explained as being part of a monotonousshale background trend. In keeping with the earlierassessment, this statistical anomaly is likewise ascribedto an edge in the data caused by fluids. Nevertheless,this pattern could hypothetically still be caused bydifferences in lithology and porosity (compaction ver-sus overpressure) and thus spurs further analysis.

Detrending elastic data for compaction effectsAlthough compaction curves vary slightly with lithol-

ogy, revisiting elastic attribute responses as a functionof depth (or using two-way traveltime as a proxy) canprove valuable to further high-grade a prospect from alead (Figure 8). For the display shown, we normalizemultiple extracted elastic seismic attributes (P-waveimpedance, Shear wave impedance, VP∕V−1

S , LambdaRho, Mu Rho, Poisson’s ratio) together with scaled seis-mic two-way traveltime. The rationale behind this ap-proach is to quickly detrend the elastic attributes forthe compaction effect because, as the dip line reveals,the target horizon crosses more than 700 ms of relief. Ascan readily be seen (from left to right), most attributesaccompany the monotonous decrease in depth with adecrease in elastic values. However, a divergence fromparalleling the background trend occurs near seismictrace 525. Here, all attributes sensitive to fluids (Pois-son’s ratio, VP/VS, and Lambda Rho) undergo a signifi-cant drop against background with Lambda Rhoattribute values being reduced by approximately 23%.Updip of this location, Lambda Rho attribute values ap-pear to be insensitive to depth change albeit at a newlylowered base level. This observation is compatible withthe Lambda Rho attribute measuring a change in elasticresponse due to the presence of a more compressiblefluid type (e.g., Young and Tatham, 2007). Notably,the Mu Rho attribute (the rigidity attribute; e.g., Chopraand Pruden, 2003) experiences a similar, but more tran-sitional drop farther downdip, approximately at seismictrace no. 410. This observed downdip transition in MuRho is interpreted as a sorting trend that incorporates

more sand-rich turbidite facies updip and graduallytransitions into silt-prone lithologies and hemipelagicmudrocks downdip (e.g., Young and Tatham, 2007;Ochoa et al., 2013). This extension of a plateau of MuRho attribute values farther downdip, i.e., past the ob-served break in the Lambda Rho attribute seen fartherupdip, permits further interpretation of the latteranomaly as probably representing a genuine fluid signa-ture. Nevertheless, for further validation, this interpreta-tion must be tested (see below) by recalibration ofseismic elasticity to a rock physics template of fluid-substituted and upscaled elastic log data from a nearbypilot well with a depth of investigation comparable tothat of the potential reservoir itself.

QI using rock physicsRock physics calibration of elastic responses of seis-

mic attributes forms one of the pinnacles of QI, becausethe fidelity of the inversion results is now known fromearlier quality control steps. Because the fidelity ofthe seismic inversion results is considered high inthis example, analytical comparison of turbidite sandsin confined canyon-fill (middle to lower slope channels)to fluid-substituted sands from an analog well — ap-proximately on depth and penetrating age-equivalentsection of similar clastic provenance — becomes pos-sible. The QI process is initiated by mapping a horizonconsisting of a single seismic line down the canyon axis(Figure 9). Lambda Rho and Mu Rho attribute valuesare then extracted along this traverse and plotted asa function of distance (Figure 9). Univariate analysisreveals that all interpreted oil-bearing section is con-tained within �2σ standard deviations. Section plottingoutboard of these bands is interpreted as wet (down-dip) or potentially gas-bearing (updip), although theseismic data ends in an updip direction. However,gas production is established from age-equivalent sandsupdip within the same petroleum system. Projection ofthe standard deviation bands into the neighboring cal-ibration crossplot confirms that the analyzed data over-lap with the stability cluster for oil-substituted sands(right panel in Figure 9e). Unfortunately, unlike the

Figure 7. Lambda Rho extraction [GPa�gcm−3] against profile distance (a) andLambda Rho elastic attribute seismic section(b). Bivariate statistical analysis supportspresence of a geobody outboard of three neg-ative standard deviations (–3σ) which corre-spond to the half-width of a Gaussiandistribution.

Interpretation / February 2014 SA133

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Figure 8. Normalized elastic attribute responses versus two-way traveltime(s).

Figure 9. Elastic calibration of seismic attribute responses to upscaled, fluid-substituted Senonian sands.

SA134 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Lambda Rho extraction, application of the same work-flow to the Mu Rho attribute extraction does not pro-duce discrimination between sand, silt, shale, ormarls known to possibly have been deposited withinthis system from regional geologic assessments. Thisis where qualitative interpretation boosts QI: the inter-preter has selected the traverse to cut one of severalprominent meandering channel features on the associ-ated base map characterized by the lowest Lambda Rhovalues. By running a second perpendicular traverse, abiconvex geobody of low Lambda Rho values can beshown to exist at the target level. The slightly biconvexmorphology of this geobody indicates differential com-paction of shale around a sand body (e.g., Chopra andMarfurt, 2012). Moreover, an interpreter may betempted to ascribe the steeply dipping elements of this

geobody to sand dikes of varying dips that pass updipinto a flattened-out tip sill (e.g., Jackson et al. [2011],their Figures 1[ii] and 8). This additional interpretationof a wing-like clastic intrusion complex (e.g., Jacksonet al., 2011) would be compatible with the tectonic set-ting of this reservoir sand being located within therealm of a transpressional conjugate margin, and theadded geologic constraint of turbidite deposition ofslope channels within the confinement provided bya submarine canyon. In either case, the discoveriesfrom seismic stratigraphy and seismic geomorphology,namely that the detected geobody dominantly consistsof sand, effectively linearizes the crossplot calibration,because the rock physics evaluation is now reduced toan analysis of the dynamic range of the Lambda Rhoattribute only. For final analysis, the interpreter cannow proceed and overlay the extracted seismic samplesfrom the traverse with the calibration crossplot. Thisstep demonstrates that the extracted attribute valuesform a cluster that dominantly overlaps with thefluid-substituted model for oil. Armed with this confir-mation, the interpreter can now proceed to map facieswithin and around the prospect. Four different seis-mic facies are defined in a crossplot of Lambda Rhovalues against elevation (Figure 10). Notably, the 19[GPa � gcm−3] contour designating the reservoir transi-tion from oil charge to the wet case conforms with aclosing time structure contour (not shown here). Inter-estingly, the Lambda Rho contour line protrudes in thestructural and stratigraphic downdip direction, sug-gesting the presence of a rigid, convex bulge stemmingfrom the differential compaction of sand surrounded byshale (in a shale-filled canyon, the contours would gen-erally point updip; Figure 11).

Fundamental concepts of seismic reservoircharacterization workflows

Wet risk, low-saturation gas, absence of reservoir,and tight reservoirs emerge as leading factors as towhy most modern clastic prospects drilled to-date fail(R. Roden, personal communication, 2013; Figure 12).

Figure 11. Seismic facies map for deepwater setting ob-tained from polygon crossplot facies in previous figure. Clos-ing contour (not shown) conforms to elastic attribute cutoffcontour (contour interval, C:I: ¼ 1 GPa � gcm−3).

Figure 12. Wet risk, low-saturation gas, absence of reservoir,and tight reservoir count among the leading causes of drywells. Display is based on 217 prospects, of which 98 provedto be dry holes (data are current through March 2012).

Figure 10. Calibrated seismic crossplot facies with fluid andlithofacies map.

Interpretation / February 2014 SA135

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

The progressive reservoir characterization workflowpresented here analyzes prognostic seismic attributeresponses that

• first, measure the probability of reservoir sandpresent,

• second, assess reservoir quality by predictingporosity, and,

• finally, predict prevailing fluid type (brine, oil,or gas).

Ultimately, visual combination of all three compo-nents in RGB displays (R: Sand, B: Porosity, G: Pay)highlights prospective map areas or seismic volumewith white emerging as the composite color indicatingthat all three critical reservoir conditions have been met(Figure 13). This discriminate use of RGB visualizationhigh-grades hydrocarbon potential and helps locate po-tential or bypassed pay on a seismic scale. To initiatereservoir characterization in deepwater clastics, the in-terpreter needs to initially detect sands within the 3Dseismic survey data. Again, the underlying corollarybeing that changes in rock physics parameters are re-flected in seismic responses of elastic attributes.

Seismic sand detectionTo seismically predict sand from poststack or

prestack data, the interpreter can choose from a largenumber of qualitative and QI methods. Qualitative inter-pretation methods include poststack seismic attributeanalysis, such as curvature study (Chopra and Marfurt,2007) or application of shale indicator technology(Taner, 2002) to detect sand from differential compac-tion features or from seismic facies patterns observedon attribute extraction maps. In contrast, rather thanseismically detecting sand from reservoir architecture,QI focuses on reservoir characterization calibrated torock physics elastic relationships. Bivariate analysesof cardinal elastic relationships between P-wave andS-wave impedances and their mathematical products,

such as Poisson’s ratio and Lamé constants (and math-ematical constructs thereof), ultimately prove to be ofgreat value in reservoir characterization. The success ofany single technique must always be evaluated againstthe data at hand because local mineralogical, diage-netic, or other peculiar reservoir conditions, such asoverpressure or changing formation water resistivities,in addition to overlapping dynamic ranges of proper-ties, often contribute to the nonuniqueness of results,thus limiting unequivocal elastic discrimination. Non-uniqueness calls for adjunct use of rock physicstemplates and subsequent creation of probability func-tions to stochastically mitigate risk (undiscussed).

Elastic calibration to sand proves very powerful andthe interpreter can choose from a plethora of sand-detection methods. For instance, the interpreter maychoose to calibrate to modified versions of the ratioversus difference method for sand detection (e.g., Liet al., 2003; Al-Dabagh and Alkhafaf, 2011). In thismethod, high-quality reservoir sands are identified bygraphing the ratio attribute (created by division ofthe Lambda Rho attribute by the Mu Rho attribute)against the difference attribute (created by subtractingthe Mu Rho attribute from the Lambda Rho attribute).This method proves valuable because the ratio compo-nent largely eliminates the compaction effect fromthe data (F. Bolivar, personal communication, 2013),allowing comparison of elastic attribute responses inregions characterized by high structural dip, such asanticlinoriums. Another effective method for sand de-tection is found by creating a single hybrid relationalelastic seismic attribute that portrays the increasingprobability of encountering sand by mapping the dis-tance from a user-defined, rock physics-validated shaleor sand baseline trend within a cross plot of LambdaRho versus Mu Rho.

Derivation and calibration of theshale background and sand trends

Superposition of elastic data from wells of vastlydifferent total depths is helpful in determining the elas-tic behavior of baseline trends with compaction. Forinstance, to derive the shale background trend frommultiwell elastic relationships, the crossplot space ispopulated with samples that have a VCLAY greater thana sensible cutoff. Petrophysical analysis of calibrationwells typically advocates retention of samples situatedabove a 70% VCLAY threshold, although actual cutoffsmay slightly differ on a case-by-case basis. The estab-lishment of a baseline trend seeks to eliminate or min-imize the influence of destabilizing effects on trendcurve calculation by excluding atypical data from cal-culation of trend statistics. Overpressure, diagenesis,mineralogy, and fluid fill may exert controls on sandor shale baseline trend position: For instance, presenceof silty shales tends to translate the baseline towardlower Lambda Rho values as can be expected to occurin the vicinity of turbidite fairways due to an increase inquartz content (sorting). Conversely, calcareous shales

Figure 13. RGB corendered attribute display. In practice,diagnostic attributes would have been calibrated to eitherelastic thresholds or cutoffs: Red gives only sand (thresholdin relational attribute); blue gives total porosity (threshold de-termined from rock physics); green gives hydrocarbon charge(cutoff determined from rock physics). Where all three reser-voir conditions are met favorably, white emerges as thecomposite color.

SA136 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

would inevitably undergo an acoustic stiffening result-ing in higher elastic values (Pelletier and Gundersen,2005; their Figures 4 and 5). In similar fashion, presenceof other heavy minerals associated with maximumflooding surface intervals may shift the idealized base-line trend toward rocks of higher rigidity (high valuesof Mu Rho attribute), although higher TOC content re-duces rigidity through such zones. Finally, overpressurewould tend to force a point cluster to migrate alonga baseline-parallel hyperbole toward smaller elasticattribute values (i.e., toward higher porosity, Figure 2).

Once a reliable shale baseline trend has been calcu-lated from a second-order polynomial regression ofupscaled well-log elastic attribute data, this relationshipcan be used to redefine the crossplot space with a singleattribute that describes an increasing probability of sandas an increasing normal distance away from the baseline.In practice, this is accomplished first by defining abaseline trend for shale and then translating this relation-ship to graphically formulate a sand line. Ordinarily, aninterpreter would have preferred using trend analysis tosolely define a sand line from upscaled log data, but toolarge a spread of elastic samples representing sands (dueto shifts in sand provenance) may at times considerablydeteriorate the reliability of the polynomial regression onsand samples andmay require defining a solid sand trendline using different means. This can be accomplishedby first defining a shale baseline and, second, translatingthe shale line across the elastic space to calibrate theposition of the sand trend line.

Reservoir quality assessmentTo obtain an assessment of reservoir quality a seis-

mic porosity volume is created via a transform thatrelates either P-wave impedance or S-wave impedanceto total porosity. The impedance-to-porosity linear orquadratic transform may additionally be optimized byinclusion of a lithology- or mineralogy-specific term(P. Alvarez, personal communication, 2013; e.g., Han,1986). The option exists to create two porosity trans-forms: one for sand and another one for shale. Sand cut-offs can be determined from modeled rock physicstemplates (Avseth et al., 2005). Merging the two poros-ity volumes creates a composite total porosity volume.The shale (total) porosity subvolume may potentiallyserve to identify an overpressured section or evaluateseal potential, whereas the sand (total) porosity volumeis well-suited as input for another transform to estimateeffective porosity.

Fluid calibration and predictionSensitivity to fluid substitution on seismic response

is evaluated by calibration of upscaled multiple welllogs of prognostic attributes and their cardinal elasticrelations (P-wave impedance versus Poisson’s ratio,λρ versus μρ and λρ∕μρ versus λρ − μρ). Often, the λρversus μρ elastic space is also determined to be the bestfor fluid calibration (in addition to sand detection).Simultaneously, rock physics templates are created

especially to assist with lithology/fluid/porosity calibra-tion by modeling fluid conditions that may not havebeen found in any of the constituent in situ wells.

DiscussionThe dictionary definition of the verb interpret is “to

give or provide the meaning of; explain; explicate; elu-cidate: to interpret the hidden meaning of a parable 2. toconstrue or understand in a particular way: to interpreta reply as favorable” (http://dictionary.reference.com).A modification of the aforementioned example (itali-cized) applied to subsurface interpretation techniquesmight then actually read “to quantitatively interpretan acoustic signal as pay via elastic calibration or stat-istical analysis.” Such is the fundamental gist of thispaper with the added emphasis on strengthening inter-pretation of seismic inversions via attribute analysis(Barnes, 2001; Taner, 2001; Chopra and Marfurt, 2007),and long-standing qualitative and QI techniques de-scribed by various workers (Latimer et al., 2000; Chopraand Marfurt, 2008). In detail, this requires pairingselected interpretation aspects of deepwater settingsas evidenced from seismic geomorphology and seismicstratigraphy (Weimer et al., 1998; Posamentier andKolla, 2003; Contreras and Latimer, 2010) with statisti-cal analyses (e.g., Latimer et al., 2000) and rock physicscalibration (Avseth et al., 2005) under special consider-ation of geologic pitfalls and oddities, such as posedby the challenge of discriminating between volcanicsills and sand injectites in amplitude data (Werner,2003; Jackson et al., 2011). Specifically, the intendedscope of this paper has been to discuss selected aspectsof qualitative and QI techniques that facilitate

1) quality control and calibration of inversion deliv-erables,

2) fast-track data reconnaissance, and3) reservoir characterization, with special emphasis on

successful prediction of fluid type(s) in deepwatersettings.

This is achieved using a variety of elastic attributesobtained from seismic inversion (P-wave impedance,S-wave impedance, Poisson’s ratio, Lambda Rho, MuRho). However, without proper calibration of elasticrelationships exhibited by seismic attribute data in bi-variate statistical analyses, reservoir characterizationon a standalone basis, using attribute visualization only,remains challenging (Figure 1). Despite the value offeredto qualitative interpretation through a perspective viewof the Lambda Rho attribute that successfully illuminatesstratigraphic elements of the Pliocene Isongo turbiditecomplex, offshore West Africa, interpretation of a singleattribute leaves unanswered what volumetric portion ofthe featured deepwater setting could potentially be pro-ducing reservoir (Figure 1). Elastic calibration of inver-sion data to (1) detect sand, (2) assess reservoir quality,and (3) predict fluid type is mandatory to reach a satis-factory conclusion to reservoir analysis. In this paper, we

Interpretation / February 2014 SA137

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

have focused on QI techniques that facilitate correctfluid prediction in deepwater clastic settings from elasticseismic attributes.

ConclusionsSuccessful QI begins with the responsible audit of

available well log and seismic data, succeeded byproper data conditioning, and proceeds with rigorousquality control, then eventually culminates in placingelastic seismic attributes within their proper geologiccontext. We provide a comprehensive workflow tothe quantitative interpreter that considers all aforemen-tioned selected aspects of subsurface interpretation byeffectively combining long-standing and proven qualita-tive interpretation techniques (elements of seismicstratigraphy and seismic geomorphology) with qualitycontrol methods, statistical analysis, data mining, elas-tic calibration (including suggestions for detrendingelastic data for compaction effects), and rock-physics-driven, QI techniques to predict fluid type duringreservoir analysis of deepwater turbidites. Specifically,we gauge the fidelity of the interpretation by comparingcardinal elastic relationships observed in upscaled welllog crossplots with the corresponding seismic inversionattributes. Matching background trends of upscaled logcurve data and seismic samples generally confirm thatan inversion is indeed in gauge. Additionally, we exploitsystemic discrepancies in P-wave and S-wave imped-ance behavior against background and within potentialreservoir zones during data mining efforts. Elasticattribute responses along seismic transects are furtheranalyzed for geologic controls, such as compactionor overpressure. Prospective seismic data are sub-sequently subject to further statistical analysis androck-physics calibration before classifying seismicfacies for map display, possible geobody extractionsor reserve estimations within an middle to lower slopecanyon-fill turbidite channel complex.

AcknowledgmentsFirst, the authors would like to thank colleagues at

Rock Solid Images, especially Lucy MacGregor, GarethTaylor, Francisco Bolivar, Pedro Alvarez, Mark Suda,and Rohit Arora for stimulating thoughts. Additionally,Lucy MacGregor, Sven Treitel, Kush Tandon, RolfAckermann, Gareth Taylor, Zakir Hossain, and RockyRoden are kindly thanked for reviewing earlier versionsof this manuscript. In addition, the authors would liketo thank the journal’s reviewers and editors, ArthurBarnes, Gary Robinson, Kurt Marfurt, and Yonghe Sun.Various anonymous industry sponsors are thanked forgranting publication rights. Aisha Roberts aided in for-matting the manuscript for publication.

ReferencesAl-Dabagh, H. H., and S. Alkhafaf, 2011, Comparison of Kρ

and λρ in clastic rocks: A test on two wells with different

reservoir-quality stacked sands from West Africa: TheLeading Edge, 30, 986–994, doi: 10.1190/1.3640520.

Avseth, P., T. Mukerji, and G. Mavko, 2005, Quantitativeseismic interpretation: Cambridge University Press.

Bahorich, M. S., and S. L. Farmer, 1995, 3-D seismic discon-tinuity for faults and stratigraphic features: The coher-ence cube: 65th Annual International Meeting, SEG,Expanded Abstracts, 94–96.

Barnes, A., 2001, Attributes in your facies: CSEG Recorder,26, 7, 41–47.

Batzle, M., and M. H. Gardner, 2000, Lithology and fluids:Seismic models of the Brushy Canyon Formation, WestTexas, in A. H. Bouma, and C. G. Stone, eds., Fine-grained turbidite systems, AAPG Memoir 72/SEPMSpecial Publication 68, 127–142.

Beauboeuf, R. T., C. Rossen, F. B. Zelt, M. D. Sullivan, D. C.Mohrig, and D. C. Jennette, 1999, Deep water sand-stones, Brushy Canyon Formation, West Texas, FieldGuide for AAPG Hedberg Field Research Conference:Deep-Water Sandstones, Brushy Canyon Formation,AAPG.

Buland, A., and H. Omre, 2003, Bayesian linearizedAVO inversion: Geophysics, 68, 185–198, doi: 10.1190/1.1543206.

Chopra, S., and K. J. Marfurt, 2007, Seismic attributes forprospect identification and reservoir characterization:SEG Geophysical Development Series 11: 464.

Chopra, S., and K. J. Marfurt, 2008, Emerging and futuretrends in seismic attributes: The Leading Edge, 27,298–318, doi: 10.1190/1.2896620.

Chopra, S., and K. J. Marfurt, 2012, Seismic attribute ex-pression of differential compaction: CSEG Recorder,37, 4, 46–51.

Chopra, S., and D. Pruden, 2003, Multi-attribute seismicanalysis on AVO derived parameters — A case study:CSEG Recorder, 28, 7, 34–38.

Clavaud, J.-B., 2008, Intrinsic electrical anisotropy ofshale: The effect of compaction: Petrophysics, 49,243–260.

Coléou, T., F. Allo, R. Bornard, J. G. Hamman, and D. H.Caldwell, 2005, Petrophysical seismic inversion: 75thAnnual International Meeting, SEG, Expanded Ab-stracts, 1355–1359.

Connolly, P. A., 1999, Elastic impedance: The LeadingEdge, 18, 438–452, doi: 10.1190/1.1438307.

Contreras, A. J., and R. B. Latimer, 2010, Acoustic imped-ance as a sequence stratigraphic tool in structurallycomplex deepwater settings: The Leading Edge, 29,1072–1082, doi: 10.1190/1.3485768.

Ehrenberg, S. N., 2004, Factors controlling porosity inUpper Carboniferous-Lower Permian carbonate strataof the Barents Sea: AAPG Bulletin, 88, no. 12, 1653–1676, doi: 10.1306/07190403124.

Gardener, M. H., and J. M. Borer, 2000, Submarine channelarchitecture along a slope to basin profile, Brushy Can-yon Formation, west Texas, in A. H. Bouma, and C. G.Stone, eds., Fine-grained turbidite systems: AAPG

SA138 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Memoir 72/Society of Economic Paleontologists andMineralogists Special Publication, 68, 195–214.

Goodway, B., T. Chen, and J. Downton, 1997, ImprovedAVO fluid detection and lithology discrimination usingLamé petrophysical parameters; “λμ”, “μρ”, & “λ∕μ fluidstack”, from P and S inversions: 67th Annual Inter-national Meeting, SEG, Expanded Abstracts, 183–186.

Han, D.-H., 1986, Effect of porosity and clay content onacoustic properties of sandstones and unconsolidatedsediments: Ph.D. dissertation, Stanford University.

Jackson, C. A.-L., M. Huuse, and G. P. Barber, 2011, Geom-etry of winglike clastic intrusions adjacent to a deep-water channel complex: Implications for hydrocarbonexploration and production: AAPG Bulletin, 95, no. 4,559–584, doi: 10.1306/09131009157.

Lancaster, S., and D. Whitcombe, 2000, Fast-track ‘col-oured’ inversion: 70th Annual International Meeting,SEG, Expanded Abstracts, 1572–1575.

Latimer, R., S. Bahret, C. Sullivan, R. Horine, W. Mills, andV. Sturrock, 2000, Reservoir characterization using geo-statistical inversion for the Amberjack Field, offshoreGulf of Mexico: 70th Annual International Meeting,SEG, Expanded Abstracts, 1506–1508.

Li, Y., J. Downton, and B. Goodway, 2003, Recent applica-tions of AVO to carbonate reservoirs in the WesternCanadian Sedimentary Basin: The Leading Edge, 22,670–674, doi: 10.1190/1.1599694.

Ochoa, J., J. Wolak, and M. H. Gardner, 2013, Recognitioncriteria for distinguishing between hemipelagic andpelagic mudrocks in the characterization of deep-waterreservoir heterogeneity: AAPG Bulletin, 97, no. 10,1785–1803, doi: 10.1306/04221312086.

Oldenburg, D. W., and R. G. Ellis, 1991, Inversion of geo-physical data using an approximate inverse mapping:Geophysical Journal International, 105, 325–353, doi:10.1111/j.1365-246X.1991.tb06717.x.

Parker, R. L., 1994, Geophysical inverse theory: PrincetonUniversity Press.

Pedersen, J. M., 1990, Simulated annealing and finite-time thermodynamics: Ph.D. dissertation, Universityof Copenhagen.

Pelletier, H., and J. Gunderson, 2005, Application of rockphysics to an exploration play: A carbonate case studyfrom the Brazeau River 3D: The Leading Edge, 24, 516–519, doi: 10.1190/1.1926810.

Posamentier, H. W., and V. Kolla, 2003, Seismic geomor-phology and stratigraphy of depositional elements indeep-water settings: Journal of Sedimentary Research,73, 367–388, doi: 10.1306/111302730367.

Russell, B. H., K. Hedlin, F. J. Hilterman, and L. L. Lines,2003, Fluid-property discrimination with AVO: A Biot-Gassmann perspective: Geophysics, 68, 29–39, doi: 10.1190/1.1543192.

Singleton, S., 2009, The effects of seismic data conditioningon prestack simultaneous impedance inversion: TheLeading Edge, 28, 772–781, doi: 10.1190/1.3167776.

Steckler, M. S., and A. B. Watts, 1978, Subsidence of theAtlantic type continental margin off New York: Earthand Planetary Science Letters, 41, 1–13, doi: 10.1016/0012-821X(78)90036-5.

Taner, M. T., 2001, Seismic attributes: CSEG Recorder, 26,no. 7, 49–56.

Taner, M. T., 2002, System for estimating the locations ofshaley sub-surface formations: U.S. Patent 6,487,502.

Tonellot, T., D. Macé, and V. Richard, 2001, Joint strati-graphic inversion of limited angle-stacks: 71st AnnualInternational Meeting, SEG, Expanded Abstracts,227–230.

Walls, J., J. Dvorkin, and M. Carr, 2004, Well logs and rockphysics in seismic reservoir characterization: OffshoreTechnology Conference, Paper Number 16921.

Weimer, P., 1998, Evaluating the petroleum systems forthe Northern Deep Gulf of Mexico through integratedbasin analysis: An Overview: AAPG Bulletin, 82, no. 5B,865.

Werner, P., 2003, Pitfalls in interpretation: Volcanics fromseismic images: 65th Annual International Conferenceand Exhibition, EAGE, Extended Abstracts, 1–4.

Whitcombe, D. N., P. A. Connolly, R. L. Reagan, and T. C.Redshaw, 2002, Extended elastic impedance for fluidand lithology prediction: Geophysics, 67, 63–67, doi:10.1190/1.1451337.

Widess, M. B., 1973, How thin is a bed?: Geophysics, 38,1176–1180, doi: 10.1190/1.1440403.

Yilmaz, Ö., 2001, Seismic Data Analysis, Vol. 1, 2nd ed.,Investigations in Geophysics No. 10: SEG.

Young, K. T., and R. H. Tatham, 2007, Lambda-mu-rhoinversion as a fluid and lithology discriminator in theColumbus Basin, Offshore Trinidad: The Leading Edge,26, 1508–1515, doi: 10.1190/1.2821936.

Uwe Strecker received a Ph.D.(1996) in geology from the Universityof Wyoming. He is a technical advisorfor Geology and Interpretation atRock Solid Images. He previouslyworked as a geologist for BelcoEnergy Corp. (1999–2001) and UPR(1997–1998). His research interestsinclude apatite fission-track thermo-

chronology and shale gas resource plays. In 2013, he wasinvited by AAPG to lead a webinar on QI of deepwaterturbidites. He has given more than 200 technical presenta-tions on seismic reservoir characterization to energyaudiences worldwide. Additionally, he has contributedto DOE reports that discuss quantification of naturalresources from cumulative seismic attributes as appliedto onshore and offshore gas hydrate deposits. He has beenan invited speaker at Marathon Petroleum TechnologyResearch Institute, Zentralinstitut für Physik der Erde,Meridian Oil Company, and Seismic Micro-Technology.He is a member of SEG, AAPG, GSH, HGS, and GSA.

Interpretation / February 2014 SA139

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/

Paola Vera de Newton received apetroleum engineering degree (2002)from the University of Zulia, Ven-ezuela, and that same year joinedVenezuelan state company PDVSAas a petroleum engineer in Zulia State.She is the Technical Advisor for Petro-physics and Rock Physics at RockSolid Images (RSI). She joined RSI

in 2007, and since then she has mainly worked in the seis-mic petrophysicist group. She has been involved in multi-ple conventional and unconventional studies worldwide.Some of her recent projects include West Africa, WillistonBasin, Gulf Of Mexico, Marcellus, Monterey, and Eagle

Ford shales rock physics studies. She is a member ofSEG, AAPG, EAGE, and SPWLA.

Maggie Smith received a B.S. (1974)in geology from the University of Glas-gow, Scotland. She began her careerin seismic data processing with geo-physical service companies, most no-tably Digicon Geophysical, and laterdeveloped an interest in combiningdisciplines for reservoir characteriza-tion studies. She has been with Rock

Solid Images since 1999.

SA140 Interpretation / February 2014

Dow

nloa

ded

08/1

4/14

to 2

16.1

98.8

5.26

. Red

istr

ibut

ion

subj

ect t

o SE

G li

cens

e or

cop

yrig

ht; s

ee T

erm

s of

Use

at h

ttp://

libra

ry.s

eg.o

rg/