Attribute expression of channel forms in a hybridcarbonate turbidite formation

Bradley C. Wallet1

ABSTRACT

Much of the world’s conventional oil and gas production comes from turbidite systems. Interpreting them inthree dimensions using commercially available software generally requires seismic attributes. Hybrid carbonateturbidite systems are an interesting phenomenon that is not fully understood. I have examined the attributeexpression of channel forms in a hybrid carbonate turbidite system from off the coast of Western Australia.I have determined several characteristic responses to attributes that improve the ability to identify and delineatethe channel forms. Finally, I have evaluated and developed a workflow that is effective at modeling andextracting the channel forms in three dimensions, leading to a product that can be used in further understandingof how carbonate turbidite systems develop.

IntroductionDeepwater marine depositional systems are one of

the most important reservoirs and have given rise tomost of the recent conventional discoveries. Turbiditesystems constitute an important class of hydrocarbonreservoirs in deepwater settings (Slatt, 2014), and con-siderable work has been published on the characteriza-tion of deepwater turbidite reservoirs.

Because of their resemblance to common earthfeatures, such as rivers, canyons, and gullies and theirstrong laterally 2D (though often stacked) nature, deep-water turbidites have inspired significant study usingseismic geomorphology. Posamentier (2006) completesa detailed analysis of a marine depositional system,looking at architectural elements at shallow depths aswell as regions of exploration interest. In his work, heanalyzes vertical, time, and horizon slices of amplitudesto illustrate channels and other elements. Additionally,he shows the value of using seismic attributes such asdip and curvature.

Braccini and Adeyemi (2011) focus upon the use ofseismic attributes to map a deepwater turbidite in off-shore Nigeria. They use spectral components, near- andfar-offset limited stack amplitudes, and coherence tovisualize channels and mass-transport complexes. Theyshow a number of vertical slices of the seismic data,demonstrating the 3D nature of these elements, but theyonly apply the attributes in map views to visualize thelateral extent of these features.

Although classic deepwater turbidites have beenthe subject of considerable study (Deptuck, 2003; Pos-amentier, 2006; Braccini and Adeyemi, 2011), signifi-

cantly less work has been done in the area of hybridcarbonate turbidite reservoirs (Weimer and Slatt, 2006).Carbonates can make an excellent environment for thedevelopment of a turbidite system. This is due to theirgrain-rich facies that can be transported by turbidityflows (Harris and Wright, 1998). Producing reservoirsof this type include the Hasa field in Abu Dhabi, theFateh field in Dubai, and the Poza Rica field in Mexico.

Another example of a hybrid carbonate turbiditeis the Mandu Formation, off the northwest coast ofAustralia. Figure 1 shows a vertical slice through seis-mic data acquired in this region taken parallel to thepaleoshoreface. Evidence of channel forms is clearlyseen in this vertical slice in the form of many lensshaped structures, and even a novice interpreter couldinterpret many though not all of these forms. Figure 2shows the same slice with a partial interpretation over-laid on the seismic.

Figure 3 shows a vertical slice perpendicular to thepaleoshoreface. The most evident feature visible in thisimage is a series of clinoforms. An experienced inter-preter can relate the internal structure of these to globaleustatic changes in sea level (Figure 4). These clino-forms constitute the Mandu Formation, an Oligoceneaged prograding carbonate shelf (Heath and Apthorpe,1984). This image gives little information useful formapping the architectural elements of the turbiditesystem.

In this paper, I evaluate the attribute expression ofa series of channel forms in the prograding carbonateMandu Formation with the goal of visualizing thechannel-form geometry in 3D, as well as in conventional

1University of Oklahoma, ConocoPhillips School of Geology and Geophysics, Norman, Oklahoma, USA. E-mail: bwallet@ou.edu.Manuscript received by the Editor 6 July 2015; revised manuscript received 19 November 2015; published online 19 April 2016. This paper

appears in Interpretation, Vol. 4, No. 2 (May 2016); p. SE75–SE86, 38 FIGS.http://dx.doi.org/10.1190/INT-2015-0108.1. © 2016 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

t

Technical papers

Interpretation / May 2016 SE75Interpretation / May 2016 SE75

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

methods. Then, I describe a workflow based for extract-ing channel forms in 3D using curvature.

Geologic and geophysical backgroundRegional geology

The study area is located on the Rankin Platform ofthe Carnarvon Basin off the coast of Western Australia(Figure 5). A major low-stand event occurred duringEarly Oligocene time. This resulted in a progradationalwedge of carbonate slope that resulted in the ManduFormation. This progradation continued until it was ter-minated by a eustatic sea-level rise approximately 15MYA (Richardson, 2000).

During the period of deposition from the Oligoceneto the middle Miocene, there were several high-stand,low-stand, and transgressive intervals, although theoverall pattern was a lower frequency sea-level rise(Figure 6). During the Oligocene to the middle Miocene,there developed a series of channel forms or canyons(for a detailed look at the geology of the Mandu Forma-tion, see Tellez Rodriguez, 2015).

There is some uncertainty as to the mechanism bywhich these form. It is believed that during low-standevents, clastic systems tend to shed the majority of

their material into the basin (Vail et al., 1977; Figure 7).However, it is unknown if this is the case in carbonateenvironments, and this topic has been the subject of con-siderable controversy (Bernet et al., 2000). Therefore,improved mapping of these canyons might add to theunderstanding of the nature of carbonates in general.

There exist a number of modern analogs for theMandu Formation. One such analog is the Little BahamaBank (Mullins et al., 1984). In this case, foreslope reefsin this zone exhibit a channel-form architecture that ispresent in my study area (Figure 8).

Another modern analog can be found in the FloridaKeys, Florida, USA. This system is a carbonate structurethat is currently experiencing subaerial exposure. View-ing this structure in aerial imagery (Figure 9) demon-strates channel forms that are very similar in theirsize and geometry to those presented in my data set.

Data descriptionThe Rosie 3D survey used in this study was provided

by Geoscience Australia (Cathro et al. 2003; Figure 10).

Figure 1. A vertical slice through the seismic data volumetaken parallel to the paleoshoreface. Evidence of channelforms is clearly seen in this image in the form of many lensshaped structures.

Figure 2. A partial interpretation overlayed on the verticalslice. The green arrows denote the channel forms in the pro-grading carbonates.

Figure 3. A vertical slice through the seismic data volumetaken perpendicular to the paleoshoreface. An experiencedinterpreter can glean considerable information regardingglobal eustatic sea levels from this view, but it does not clearlymap the channel forms and other architectural elements of theturbidite system.

Figure 4. A vertical slice through the seismic data volumetaken perpendicular to the paleoshoreface with an interpreta-tion overlain. The blue lines indicate the tops of high-standevents. The green lines indicate the tops of transgressiveevents. The red lines indicate the tops of low-stand events.

SE76 Interpretation / May 2016

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

It was acquired between November 1996 and February1997 by Geco-Prakla, with a line orientation of north-east–southwest (49.5°) with a dual-source/four-streamerconfiguration. The cable separation was 150 m and thesource separationwas 75m, resulting in an initial subsur-face CDP line spacing of 18.75 m. Each streamer had 304channels, resulting in a nominal fold of 50.

Processing was completed by Western Geophysicalfrom December 1996 through July 1997 with a resultingsample increment of 4 ms. The data are of good quality

and broad bandwidth for the time of acquisition andprocessing. Dominant frequency through the ManduFormation is approximately 35 Hz (Figure 11) (more de-tails about the acquisition and processing of this dataset can be found in Richardson, 2000).

Seismic expression of channel formsHistorically, interpretation has reflected the 2D ori-

gins of seismic analysis with interpreters using verticalslices through seismic data to interpret geologic features,

such as faults and horizons. By interpret-ing these features on successive slices,an interpreter is thus able to construct3D interpretations using an adapted 2Dapproach.

Geologic modeling or geomodeling isthe science of using geologic and geo-physical data to construct computerizedmodels of the subsurface. Typically, thisis thought of as being done in 3D. Al-though successive 2D interpretationscan construct 3D models of some geo-logic features, it is more efficient to usea direct 3D approach.

To establish a background for 3Dgeomodeling of the channel forms, I ex-amined the expression of channel formsin my seismic data using a number ofseismic attributes. The focus was uponvisualizing these channel forms usingseismic geomorphological methods com-bined with seismic attributes. Recently,there has been a trend in published workto focus upon the attribute expression ofspecific geologic features. These worksexamine a suite of attributes, and showexamples of how these features are illu-

Figure 5. Location of the Carnarvon Basin along the coast and offshore ofWestern Australia. ©Commonwealth of Australia (Geoscience Australia). Usedwith permission.

Figure 6. Relative sea-level curve for the Rankin Platform from the Oligocene to the present.

Interpretation / May 2016 SE77

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

minated with various attributes. The goalof this work is to illustrate how specificattributes are tied geologically and geo-physically to the feature of interest.

This approach is in contrast to anall too common approach to seismicattributes of calculating a large numberof attributes with little regard for theirunderlying geologic or geophysical tieto the interpretation task at hand. Com-mercial packages typically have 50+attributes, many of which are poorlyunderstood by interpreters. The problemis compounded when the interpreterthen relies upon partially understooddata analysis techniques, such as neuralnetworks or latent space modeling (Wal-let et al., 2009) to automatically processthese attributes without careful exami-nation.

Figure 7. Bypass margins are believed to form turbidity flows during low-standevents. (a) Transgressive-stand event, (b) high-stand event, (c) low-stand event,and (d) drowning unconformity.

Figure 8. One modern analog for the Mandu Formation canbe found on the Little Bahama Bank. Bypass margin slopes inthis zone exhibit a similar channel-form architecture.

Figure 9. A modern analog for the Mandu Formation that iscurrently experiencing subaerial exposure is the Florida Keys,Florida, USA. Similar channel-form architecture is readily vis-ible in aerial images (Courtesy of Google Earth™).

Figure 10. The location of the Rosie 3D survey that is used inthis study.

Figure 11. Frequency spectrum of the Rosie 3D data setthrough the Mandu Formation.

SE78 Interpretation / May 2016

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

In contrast, interpreters who are experienced in theuse of seismic attributes typically use a very small num-ber of attributes they understand well, and that theyhave found useful in the past. However, the ability todo this properly involves understanding and experienceas to what attributes work within a given geologic anddata quality setting. Furthermore, we are happiest whenwe can link the response of a chosen attribute to somegeologic and/or geophysical feature of our data, andthus explain our results in a meaningful manner. Withthis in mind, I will test a number of attributes, focusingupon their expression of channel forms in the hybridcarbonate turbidite system.

Single attribute analysisI begin my analysis by examining individual attributes

to better understand the seismic attribute expression ofthe channel forms. Close examination of Figure 1 sug-gests three distinct sets of channel forms. Approximatelocations of these sets of channel forms are designated inFigure 12.

Because seismic geomorphology involves using visu-alization techniques to illuminate features at a givengeologic time, a common workflow is to pick horizonslices and flatten the data rather than work with time

slices. However, as seen in Figure 2, more than halfthe lateral extent of the Mandu Formation is channelforms. In this environment, picking a consistent surfacefor flattening proved impossible. Best practice in thiscase typically involves picking a strong continuous re-flector above or below the channelized region and gen-erating phantom horizon slices (Brown, 2011). Figure 13shows a horizon I interpreted for this purpose.

Figure 14 shows the results of flattening using thehorizon shown in Figure 13. This appears to have flat-tened the data well in the lowest portion of the ManduFormation. However, given the strong progradationalnature of the formation, the flattening horizon has littlebearing upon the internal geometry of the Mandu For-mation. Hence, further up in the system, the flatteninghas little result, and horizon slices will tend to cut acrossgeologic ages, which should be recognized as a limitationof this flattening workflow.

Figure 15 shows a magnified view of the vertical sliceof amplitude data centered about a representative chan-nel form. Several details concerning the channel form

Figure 12. Three distinct sets of channel forms are apparentin the data set. Approximate locations of these sets are des-ignated by the green, red, and blue sections.

Figure 14. This is the flattened version of the data that wasused in this study.

Figure 13. The green line shows the horizon I interpreted toflatten the data volume for this study.

Figure 15. Magnifying a channel form yields considerable in-formation about the morphology of the features. Reflectors as-sociated with the channel forms tend to converge out towardthe edges inside the channel forms (blue arrow) and in towardthe edges outside the channel forms (red arrow). Also, thereare considerable numbers of discontinuities associated withthe channel forms (green arrow). The overall shape of the chan-nel forms is concave as shown by the yellow line.

Interpretation / May 2016 SE79

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

suggest attributes that may be useful. Specifically, theshown channel form is associated with discontinuities,converging reflectors, and an overall concave shape.

Figure 16 shows a horizon slice taken 52 ms abovethe flattening surface. Many meandering channel formsare visible in this horizon, which appear to have beendeposited upon a relatively flat topography. My inter-pretation is that these systems formed in a depositionalenvironment before the carbonate reef prograded intothe region of this data set.

Figures 17 and 18 show horizon slices taken 152 and352 ms above the flattening horizon, respectively. Thechannel forms in these horizons cut sharply throughthe carbonate, and they appear to result from erosionrather than deposition.

The variance attribute detects boundaries and edgesby calculating the second central moment (statisticalvariance) for a windowed region in a seismic data set(Marfurt et al., 1998). In the presence of a boundary, thewindow contains observations from either side of theboundary, thus resulting in a higher variance value.Variance would thus help in interpreting channel formsby highlighting their edges. Figure 19 shows the vari-ance attribute for a vertical line. In this image, the chan-nel forms are visible. However, the edges and form ofthe channel forms are not clear, and the overall impres-sion is “blurry.”

Figure 16. Phantom horizon slice through the seismic ampli-tude volume 52 ms above the flattening (green) horizon. Thechannel forms are meandering in this horizon. I interpret thisas being before the carbonate reef prograded into the region.

Figure 17. Phantom horizon slice through the seismic ampli-tude volume 152 ms above the flattening (green) horizon. Thechannel forms are straighter and wider than they were at theshallower horizon corresponding to their deeper incisement.

Figure 18. Phantom horizon slice through the seismic ampli-tude volume 352 ms above the flattening (green) horizon. Theextent and edges channel forms are more difficult to interpretin this horizon because they are cutting through the internalgeometry of the carbonates at a much higher angle.

Figure 19. The variance attribute for a vertical line. The ex-pression of the channels is readily apparent, but edges are notclear. The image has a generally blurry appearance.

SE80 Interpretation / May 2016

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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 20 shows a magnified view of the varianceattribute corendered with the seismic amplitude over achannel form. Although variance detected some of theboundaries, it missed others, and it did not provide for aclosed boundary.

Figures 21, 22, and 23 show the variance attribute for52, 152, and 352 ms above the flattening (green) horizonshown in Figure 13. Unlike in the vertical slice (Figure 19),the edges are clear and crisp. This contrast suggeststhat although variance may be useful in understandingand interpreting channel forms, it is not suitable for 3Dgeomodeling of them as it does not achieve closure(fully defining the boundaries in 3D).

The next attributes I will consider is the principal cur-vature attribute: most-negative principal (k2) curvature(Mai et al., 2008). Because channels are frequently de-scribed as “lens-shaped” or valley cross-sectional struc-tures in seismic data, it seems reasonable that channelforms would have a curvature response, and hence prin-cipal curvature should be useful in their interpretation.

Highly negative values of k2 would tend to be asso-ciated with structural valleys. Generally, this is theexpected shape of seismically thick channel forms.

Figure 20. The variance attribute magnified in upon a chan-nel form. Some boundaries are highlighted (green arrows),whereas others are not (red arrow). The bottom of the chan-nel form, however, is not highlighted (black arrow).

Figure 21. Variance attribute for 52 ms above the flatteninghorizon.

Figure 22. Variance attribute for 152 ms above the flatteninghorizon.

Figure 23. Variance attribute for 352 ms above the flatteninghorizon.

Interpretation / May 2016 SE81

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

Figures 24 and 25 shows the k2 curvature corenderedwith seismic amplitude for the vertical slice. In this fig-ure, the highly negative values of k2 (dark blue) areclearly associated with the channel forms. However,these regions do not appear to span the full widthof the channels but rather accentuate the channel-formaxes.

Figures 26, 27, and 28 show the k2 curvature attrib-ute for 52, 152, and 352 ms above the flattening (green)horizon shown in Figure 13. In Figures 26–28, the highlynegative regions appear to completely map the chan-nels. Furthermore, unlike variance, they appear to cor-respond to the channel forms axes rather than the

Figure 24. The k2 principal curvature for a vertical. Highlynegative values appear to follow the centers of the channelforms, but they do not appear to fully cover the lateral cross-wise extent.

Figure 26. The k2 principal curvature for 52 ms above theflattening horizon. Highly negative values appear to map outthe channel forms.

Figure 27. The k2 principal curvature for 152 ms above theflattening horizon. Highly negative values appear to map outthe channel forms.

Figure 25. Most-negative curvature magnified in upon achannel form. Highly negative values of k2 characterize thevertical extend of the channel form, they do not reach thefull-lateral extent as shown by the green arrows.

Figure 28. The k2 principal curvature for 352 ms above theflattening horizon. Highly negative values appear to map outthe channel forms.

SE82 Interpretation / May 2016

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

channel-form boundaries. This makes them suitable forgeobody extraction methods based upon value thresh-olding. However, I note that Figure 24 suggested thatsuch an approach might not lead to the modeling ofthe full-crosswise extent of the channel forms.

Because k2 curvature appears to map the center ofthe channel forms, I turn my attention to what might beused to “fill in the gaps.” Examination of Figure 15 leadsto the observation that although the middle of thechannel forms have a trough-like appearance, theedges “roll-off” into angular unconform-ities. Reflector convergence (Marfurtand Rich, 2010) is an attribute sensitiveto such unconformities. It works by cal-culating the magnitude to which reflec-tors converge together, as well as thedirection of any such convergence.

Figure 29 shows the reflector conver-gence azimuth and magnitude for thevertical line. The sides of the channelforms appear visible as adjoining re-gions of values separated by 180°. Ingeneral, the values of these regions aremostly in east/west sets perpendicularto the channel-form axes. Contrastingvalues appear on the flanks of the chan-nel forms. These values on the flanks arecaused by the erosional nature of thechannel forms’ cross-cutting layers.

Figure 30 shows a magnified portionof the vertical line. Contrasting azimuthsare seen on the edges of the channel formsand the flanks. Figure 31 shows how sucha configuration will tend to occur in an in-cised channel-form architecture.

Geobody extractionGiven these observations on the re-

sponse of the k2 curvature attribute tothe channel forms, I can now use athresholding approach to modeling andextracting the channel forms in 3D (Myeret al., 2001). Geobody extraction is aprocess by which attribute values in oneor more 3D volumes are used to createobjects in 3D. These objects could corre-spond to salt diapers, channels, or othergeologically meaningful features. Fig-ure 32 summarizes typical workflow usedfor geobody extraction (Moore, 2001).

The first step in the geobody extraction workflow isthe selection of attribute(s). In the previous section, Ilooked at a number of attributes and I concluded thatk2 curvature was the best candidate.

The next step in the geobody extraction workflow isto select attribute parameters. Although k2 curvaturewas a good candidate for geobody extraction, highly neg-ative values did not extend fully across the lateral extentof the channel forms. I took measurements of a number

of channel forms, and determined that 600 m was atypical lateral width for the channel forms. I then con-structed a spatial operator with a half wavelength of600 m to use in the curvature calculation process. Thisoperator is shown in Figure 33.

Using the spatial operator shown in Figure 34, Icalculated k2 curvature for the extent of the ManduFormation (Figure 34). In this case, the highly negativevalues did a much better job extending across the fulllateral extent of the channel forms.

Figure 29. Reflector convergence azimuth and magnitude for a vertical line.The reflectors converge on the edges of the channel forms and the flanks.

Figure 30. Reflector convergence azimuth and magnitude for a magnified por-tion of a vertical line. The green arrows denote the convergence associated withthe edges of the channel forms, and the red arrows show the convergence as-sociated with the flanks.

Figure 31. Reflectors on the flanks of an incised channelform will tend to converge toward the edges.

Interpretation / May 2016 SE83

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

Using the calculated k2 curvature shown in Figure 34,I then established a threshold to be used in the geobodyextraction process (Figure 35).

Figure 36 shows the values that fell below the se-lected threshold. These results appear to well define thechannel forms and were thus used to continue the geo-body extraction workflow.

The final step of the geobody workflow involvespostprocessing of the thresholded values. This step in-cludes region growing to define objects and a postpro-cessing step to reject objects that are too small in size.The results of this extraction process are shown in Fig-ure 37. The extracted channel forms appear to be geo-logically feasible, and correlate closely to the channelforms seen by animating through a suite of east–westvertical slices through the seismic amplitude data. Thisseems to confirm the viability of the use of k2 curvaturefor 3D geomodeling.

Figure 38 shows a representative intersection of theextracted geobodies with the seismic amplitude for thevertical line (Figure 14) used in the validation process.The geobodies in this slice have modeled every channel

Figure 33. A spatial operator with a half-wavelength of600 m.

Figure 34. The k2 curvature calculated using a spatial oper-ator designed for 600 m wide features calculated for the ex-tent of the Mandu Formation. Highly negative values of k2 doa much better job of extending across the full-lateral extent ofthe channel forms. Note that several of the “interfluvials” orregions between the channel forms have a dome shape, whichgives rise to a positive value (red) for k2.

Figure 35. In the thresholding process, highly negative val-ues of k2 curvature are assumed to correspond with the chan-nel forms.

Figure 36. Values of k2 curvature below the selected thresh-old are candidates for inclusion in the geobody being ex-tracted.

Figure 32. A typical geobody extraction workflow.

SE84 Interpretation / May 2016

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

form that is present in the data. However, there weresome additional small valley features that were ex-tracted below the larger channel forms that should nothave been. Furthermore, as expected, the modeled geo-bodies do not cover the full crosswise extent of thechannel forms. The overall results are quite promising,although further parameter tuning is necessary.

ConclusionsIn this chapter, I have examined the seismic attribute

expression of channel forms in a hybrid carbonate tur-bidite formation using a 3D data set from the CarnarvonBasin, Australia. This formation is highly channelized,and mapping these channels through traditional seismicgeomorphological methods is complicated by the diffi-culty in picking appropriate flattening horizons. There-fore, I have designed and implemented a more modernworkflow based on seismic attributes.

I have shown how these channel forms are expressedby a number of seismic attributes including variance,

principal curvature, and reflector convergence. Thesewere picked due to articulable characteristics of themorphology of channels and channel-like structures. Ialso used multiattribute visualization to demonstratethe interaction and synergy of multiple attributes.

In my analysis, I found that k2 curvature was veryeffective in mapping the trough-like nature of the chan-nel forms. The utility of k2 was enhanced by the factthat the channel forms were seismically thick relativeto velocity differences. I then used this relationshipto build 3D models of the channel forms. I showed thatthis approach provided promising results in that it ap-pears to map all of the channel forms. However, it didhave a weakness in that it captured a small number ofunwanted features. Furthermore, although it mappedthe channel forms, it did not provide complete coverageof their crosswise extent.

I have laid the groundwork for future work. Specifi-cally, further tests could be used to reject falsely map-ped regions. Additionally, I would suggest using the ex-tracted channel-form models as a starting point forheuristic analysis that would expand the models untilthe full extent was captured.

Carbonate turbidite systems are not as widely under-stood as the more traditional clastic turbidites, andthere exist questions concerning how they are formed.Specifically, it is unknown during what timing of thesystem evolution they are formed. Accurate and com-plete modeling of these channel forms should help tofurther study their nature, contributing to our under-standing of carbonate turbidite systems.

AcknowledgmentsI would like to thank Geoscience Australia for pro-

viding the data used in this paper. Additionally, I wouldlike to thank the sponsors of the Attribute Assisted Seis-mic Processing and Interpretation consortium for theirongoing support. Finally, I would like to thank K. Mar-furt and R. Slatt for their guidance and advice.

ReferencesBernet, K. H., G. P. Eberli, and A. Gilli, 2000, Turbidite

frequency and composition in the distal part of theBahamas transect: Proceedings of the Ocean DrillingProgram, Scientific Results, 45–60.

Braccini, E., and A. Adeyemi, 2011, Integrating seismicattributes for geo-modeling purposes: Nigeria deep-water turbidite environment case study: Proceedingsof the Gulf Coast Section of the Society for SedimentaryGeology.

Brown, A. R., 2011, Interpretation of three-dimensionalseismic data: AAPG.

Cathro, D. L., J. A. Austin, and G. D. Moss, 2003, Progra-dation along a deeply submerged Oligocen-Miocene het-erozoan carbonate shelf: How sensitive are clinoformsto sea level variations?: AAPG Bulletin, 87, 1547–1574,doi: 10.1306/05210300177.

Figure 37. Extracted geobodies for the channel forms mod-eled using k2 curvature. The results appear to be geologicallyfeasible.

Figure 38. Extracted geobodies overlain on the seismic am-plitude data.

Interpretation / May 2016 SE85

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/

Deptuck, M. E., 2003, Post-rift geology of the Jeanne d’Arcbasin, with a focus on the architecture and evolutionof early Paleogene submarine fans, and insights frommodern deep-water systems: Ph.D. thesis, DalhousieUniversity.

Harris, P. M., and R. P. Wright, 1998, Reservoir potential ofthe carbonate forereef-slope setting: Presented at theAAPG International Conference and Exhibition.

Heath, R. S., and M. C. Apthorpe, 1984, New forma-tion names for the Late Cretaceous and Tertiarysequence of the southern North West Shelf: TechnicalReport Record 1984/7, Geological Survey of WesternAustralia.

Mai, H. T., K. J. Marfurt, and S. Chávez-Pérez, 2008,Coherence and volumetric curvatures and their spatialrelationship to faults and folds, an example from Chic-ontepec Basin, Mexico: SEG.

Marfurt, K. J., R. L. Kirlin, S. Farmer, and M. S. Bahorich,1998, 3-D seismic attributes using a semblance-basedcoherency algorithm: Geophysics, 63, 1150–1165, doi:10.1190/1.1444415.

Marfurt, K. J., and J. Rich, 2010, Beyond curvature— Volu-metric estimates of reflector rotation and convergence:80th Annual International Meeting, SEG, ExpandedAbstracts, 1467–1472.

Moore, C. H., 2001, Carbonate reservoirs porosity evolu-tion and diagenesis in a sequence stratigraphic frame-work: Elsevier.

Mullins, H. T., K. C. Heath, H. M. V. Buren, and C. R. New-ton, 1984, Anatomy of a modern open-ocean carbonateslope: Northern Little Bahama Bank: Sedimentology,31, 141–168, doi: 10.1111/j.1365-3091.1984.tb01956.x.

Myer, D. E., E. L. Harvey, T. E. Bulloch, J. C. Voncannon,and T. M. Sheffield, 2001, Use of seismic attributes

in 3-D geovolume interpretation: The Leading Edge,20, 1377–1400, doi: 10.1190/1.1486768.

Posamentier, H. W., 2006, Seismic geomorphology: Imag-ing elements of depositional systems from shelf to deepbasin using 3D seismic data: Implications for explora-tion and development: Presented at the 3D SeismicTechnology: Application to the Exploration of Sedimen-tary Basins.

Richardson, M. J., 2000, Rosie 2D and 3D seismic interpre-tation report: Technical report, Woodside Energy Limited.

Slatt, R. M., 2014, Stratigraphic reservoir characterizationfor petroleum geologists, geophysicists, and engineers:Elsevier.

Tellez Rodriguez, J. J. J., 2015, Seismic-sequence stratigra-phy framework and architectural elements for Cenozoicstrata at the Rankin Platform sub basin, north Carnarvonbasin, Australia: Master’s thesis, University of Oklahoma.

Vail, P. R., R. M. Mitchem, and S. Thompson, 1977, Seismicstratigraphy and global changes of sea level. Part 2: Thedepositional sequence as a basic unit for stratigraphicanalysis, in C. E. Payton, ed., Seismic stratigraphy: Ap-plications to hydrocarbon exploration: AAPG Memoir165, 53–62.

Wallet, B. C., M. C. de Matos, J. T. Kwiatkowski, and Y.Suarez, 2009, Latent space modeling of seismic data: Anoverview: The Leading Edge, 28, 1454–1459, doi: 10.1190/1.3272700.

Weimer, P., and R. M. Slatt, 2006, Introduction to the petro-leum geology of deepwater settings: AAPG, AAPG Stud-ies in Geology.

A biography and photograph of the author are notavailable.

SE86 Interpretation / May 2016

Dow

nloa

ded

05/1

1/16

to 1

29.1

5.66

.178

. 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/