Earth-Science Reviews 47 Ž1999. 189–218 www.elsevier.comrlocaterearscirev
Definition of subsurface stratigraphy, structure and rock properties from 3-D seismic data Bruce S. Hart
)
New Mexico Bureau of Mines and Mineral Resources, Socorro, NM 87801, USA Received 2 February 1998; accepted 12 April 1999
Abstract This paper summarizes how three-dimensional Ž3-D. seismic technology is being used, primarily in the petroleum industry, to define subsurface structure, stratigraphy and rock properties. A 3-D seismic data volume: Ža. provides a more accurate image of the subsurface than can be obtained with 2-D seismic methods; Žb. is continuous, and so has a much greater spatial sampling than is obtained with 2-D seismic or other subsurface data Že.g., wells.; and Žc. can be viewed and interpreted interactively from a variety of perspectives, thus enhancing the interpreter’s ability to generate an accurate description of subsurface features of interest. Seismic interpretation was once the almost exclusive realm of geophysicists, however, most 3-D seismic interpretation today is conducted by multidisciplinary teams that integrate geophysical, geological, petrophysical and engineering data and concepts into the 3-D seismic interpretation. These factors, plus proper survey design, help to increase the chances of success of a 3-D seismic interpretation project. Although there are cases where the technology is not appropriate or cannot be applied Žfor economic reasons or otherwise., the general success of 3-D seismic has led it to become a mainstay of the petroleum industry. The approach and technology, first developed in that industry, have potential applications in other applied and fundamental earth science disciplines, including mining, environmental geology, structural geology and stratigraphy. q 1999 Elsevier Science B.V. All rights reserved. Keywords: seismic; petroleum; structure; stratigraphy; rock properties
1. Introduction Few technologies have affected a geoscience subdiscipline to the extent that three-dimensional Ž3-D. seismic has affected petroleum geoscience Žgeology and geophysics.. Like the field of seismic stratigraphy before it Žwhich more or less directly spawned sequence stratigraphy., 3-D seismic technology was developed in the petroleum industry and until re)
Fax: q1-505-835-6333; E-mail:
[email protected]
cently has been utilized almost entirely within that field. However, potential applications of 3-D seismic data in other geoscience disciplines are many, including studies in structural geology, stratigraphy, geophysics and petrophysics. The mining industry has begun to investigate the potential of 3-D seismic to identify and map ore bodies, and to plan mine development ŽEaton et al., 1997.. Furthermore, it may be possible to transfer the technology cost-effectively to the environmental sector ŽSiahkoohi and West, 1998.. The cost and technical requirements of collecting
0012-8252r99 r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 1 2 - 8 2 5 2 Ž 9 9 . 0 0 0 2 9 - X
190
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
and interpreting 3-D seismic data have prevented, until recently, most academic researchers from obtaining and utilizing them, although some 3-D seismic data have been collected to study deep crustal structure Že.g., Kanasewich et al., 1987, 1995. and others have been collected in conjunction with the ocean drilling program ŽShipley et al., 1994.. Changes within the petroleum industry that have been brought about by the use of 3-D seismic also have had a significant, albeit indirect, impact in other areas. One example is the use of computer workstations in data visualization. Some large 3-D seismic surveys contain gigabytes of data. Seismic interpreters have been pushing software and hardware developers to be able to visualize and interpret these enormous data sets interactively, and the hardware, software and concepts developed this way will have an impact on other earth science fields. The 3-D seismic revolution has helped to promote the development of multidisciplinary teams. Integration of seismic and geological data and concepts in workstation environments has been undertaken for at least a decade Žcf., Cross and Lessenger, 1988.. However, the continuous coverage provided by 3-D seismic data has revealed details of reservoir complexities that cannot be characterized using 2-D seismic and well control. Multidisciplinary teams in many companies integrate different data types and use the 3-D seismic workstation Žand derived products such as maps and volume interpretations. as focal points for their exploration and development efforts. In some regions and organizations, wells are not drilled without previously collecting and interpreting 3-D seismic data. Why have 3-D seismic and associated technologies had such a profound impact on the petroleum industry? What are the benefits of 3-D seismic? How does one acquire and interpret such data? This summary will address these questions, and present answers in terms that will be accessible to the geoscience community at large. The main purposes of this paper are: Ža. to illustrate how 3-D data are collected and interpreted in the petroleum industry, and Žb. to suggest how other subdisciplines of geoscience might exploit the technology. There are other developing fields of seismic technology, such as amplitude variation with offset, vertical seismic profiling, seismic inversion and cross-
well seismic technology that are often used in conjunction with 3-D data. It is beyond the scope of this paper to discuss these topics. Furthermore, a complete review of the principles of seismic surveying and interpretation will not be presented here. Instead, only those aspects that are germane to 3-D seismic collection, processing or interpretation will be presented here. Numerous other sources present information about the seismic method Že.g., Sheriff and Geldart, 1995.. Brown Ž1996a. dealt exclusively with 3-D seismic analyses, and Weimer and Davis Ž1996. presented many illustrative case studies of the use of 3-D seismic data, principally for petroleum exploration and development.
2. 3-D seismic acquisition and processing Like 2-D seismic studies, 3-D seismic data are acquired by generating an acoustic pulse at or near the surface Žland surface or sea surface., and recording the energy that is reflected from subsurface changes in physical properties Žspecifically, velocity and density.. Contacts between stratigraphic units having detectable changes in physical properties Žoften associated with the tops of formations, members, etc.. will cause reflections and be picked as seismic horizons 1 during the interpretation phase in order to discern structural and stratigraphic details of interest. On land, sources are typically vibroseis trucks or dynamite, and the receivers are geophones that detect ground motions. At sea, 3-D seismic surveys are collected using airgun arrays, with pressure sensitive hydrophones detecting the reflected energy. In almost all cases, it is compressional wave Ž‘‘p wave’’. energy that is recorded, although multicomponent 3-D surveys that include recording of shear wave reflections are becoming more common Že.g., Arestad et al., 1996.. With 2-D seismic data collection, sources and receivers are laid out along a line, and the reflected energy is assumed to come from a point mid-way between source and receiver Žthe common midpoint.
1
Italicized terms are defined in Appendix A.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
along a vertical 2-D plane. In the case of horizontally layered strata, this assumption is valid. However, in areas where there is appreciable subsurface structure, reflected energy can be recorded from interfaces that
191
do not lie mid-way between sources and receivers. Seismic migration is a processing technique that attempts to reposition this reflected energy to its true subsurface location. The need for migration of 2-D
Fig. 1. Seismic modeling results showing an example of sideswipe and Fresnel zone effects associated with a reef. The upper image shows the model in map view, illustrating the locations of seismic transects shown below. The reef is visible in a seismic transect that is over 300 m to the side. Although the sections have not been migrated, 2-D migration will not be able to remove the sideswipe image of the reef from the transects ŽJackson and Hilterman, 1979; cited by Crawley Stewart, 1995.. Reproduced with permission from Hilterman.
192
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
seismic data has been recognized from the early stages of seismic exploration ŽSheriff and Geldart, 1995.. Where subsurface structural or stratigraphic entities have a distinct three dimensionality Ži.e., many or most areas of geologic interest., reflections may be recorded from interfaces that are outside of the plane containing the source and receivers ŽFig. 1.. These reflections, sometimes called sideswipe, cannot be removed from seismic profiles using 2-D migration. It may be impossible to distinguish this sideswipe from reflections that are truly in the plane of the seismic profile, and maps or interpretations that are drawn from such seismic sections will be erroneous. Applied and theoretical examples of this problem have been presented by French Ž1974., Crawley Stewart Ž1995. and Brown Ž1996a. among others. The acquisition of 3-D seismic data exploits the spherical expansion of the acoustic pulse in the subsurface away from the source. On land, one common acquisition pattern spreads receiver groups out in lines that are oriented at 908 to the shot lines ŽFig. 2.. Reflections from each shot are recorded by many geophones, producing a row of common midpoints that is perpendicular to the orientation of the source lines, and parallel to the orientation of the receiver lines. By moving the shot location, a subsurface grid of common midpoints is generated. Note that the acquisition geometry shown in Fig. 2 is a gross simplification of real survey design. Although there are many different designs that can be used, each survey typically consists of many parallel source lines and many parallel receiver lines that are oriented perpendicular to the source lines. Each shot is recorded by a patch of geophones. In this way, individual midpoints are imaged by different combinations of sources and receivers, thus, building up the fold Žor multiplicity . of the survey. Higher fold data, all else being equal, will result in a higher signal-to-noise ratio and therefore more interpretable seismic data. Experience has shown that the fold of a 3-D survey needs only to be about one half the fold of a 2-D survey to obtain the same interpretability ŽB. Hardage, personal communication, 1994.. At sea, 3-D seismic data are generally acquired by ships towing airgun arrays and hydrophone streamers that sailed back and forth across the survey area.
Fig. 2. Sample source and receiver layout for a land-based 3-D seismic acquisition program. In this simple case, each shot is recorded by a line of receivers. By moving the source location, a rectangular grid of common midpoints is generated. In practice, many source and receiver lines would exist, increasing the number of source — receiver combinations that image a particular midpoint to build up the fold Žand data quality. of the survey. Other, more complex, survey layouts are typically used in practice.
Increasingly though, innovative techniques such as using two vessels simultaneously, or implanting geophones on the sea floor, are being developed and exploited. The spacing between the midpoints in the receiver direction is one half the distance between the re-
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
ceivers, while the spacing in the source direction is one half the distance between the source locations. For example, assuming that source and receiver locations are both at 60 m intervals, midpoints will be generated every 30 m = 30 m. In this case, each midpoint represents an area or bin of 30 m = 30 m. The bin size might be rectangular, rather than square, if the distances between source and receiver locations are not identical. By acquiring continuous coverage of seismic data in three dimensions, the data can be migrated in three dimensions. In this way, reflection energy is accurately repositioned to its true subsurface location, and transects through a 3-D seismic volume will only show those features that are truly in the plane of the section Že.g., Brown, 1996a.. As such, all else being equal, a vertical transect through a 3-D seismic volume is a more accurate, better image than an equivalent seismic transect derived using 2-D acquisition and processing.
193
of as a series of cubes, or Õoxels Ža term analogous to the 2-D pixels of remote sensing., each of which stores a particular amplitude value ŽFig. 3.. Using modern computer graphics capabilities, it is possible to visualize and interpret the seismic data in a variety of ways Žsee below.. When seismic data were primarily viewed as paper sections, the standard display format was to use black and white variable area wiggle displays. This display presents seismic traces as continuous curves that define a time series of positive Žpeaks. and negative Žtroughs. amplitudes, and the peaks are filled in with a solid black colour. A drawback of the variable area wiggle display is that the eye tends to focus on the peaks, and the
3. Viewing 3-D seismic data Virtually, all interpretation of 3-D seismic data is conducted on workstations or powerful personal computers ŽHart, 1997.. By exploring and visualizing both the data and interpretations in progress, the geoscientist can derive a much better understanding of the three dimensionality of subsurface stratigraphic and structural elements. This allows the interpreter to generate a more accurate subsurface description Ždepicted in maps, etc.., than by working with paper seismic displays andror well logs alone. These visualization technologies further aid the geoscientist in conveying the results of his or her work to others who have not been involved in the interpretation process. Each bin in a 3-D seismic volume Žhaving x, y dimensions that are defined primarily by data acquisition operations. can be represented by a single seismic trace, conceptually centered in the middle of the bin. Each trace in turn is divided into equal increments in the z direction that define the sampling interval. For petroleum exploration purposes, the sampling interval is typically 2 or 4 ms. The result is that the 3-D seismic volume can be thought
Fig. 3. Conceptual diagram showing a 3-D data volume. Each voxel in such a volume is characterized by x, y, z coordinates and an amplitude value Žpositive and negative values.. The x and y dimensions of each voxel represent the bin size, and are a function of acquisition parameters, whereas the z dimension of a voxel represents the digital sampling interval. The range of amplitudes depends upon how the data have been scaled. In a variable density display Žtop and side of the cube., the numerical amplitude values are represented by colours that are selected by the interpreter Žsee Fig. 4..
194
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
Fig. 4. Colour displays of seismic data. Ža. Seismic data has traditionally been displayed as Õariable area wiggle displays Žleft. in which the reflected energy is displayed as waveforms consisting of positive and negative amplitude values Žpeaks and troughs respectively.. Peaks are filled in with black to enhance the interpretability of the seismic transect. Computer graphics displays allow the amplitudes Žcenter. to be colour coded so that positive and negative amplitudes are shown, respectively, in blues and reds with the darkness of each colour being proportional to the amplitude value. Whites correspond to zero amplitude. By removing the wiggle trace Žright. the seismic traces are shown as Õariable density displays. Žb. Variable density display of a seismic transect. By displaying adjacent traces side by side, the display has the appearance of data continuity throughout the entire transect. Compare with the variable area wiggle display of the same transect Žc.. The colour display gives approximately equal weight to peaks and troughs Žblue and red, respectively., helping the interpreter to detect stratigraphic and structural features. Variations in amplitude along a single reflector are also easier to detect. Although the blue–white–red colour scale shown here is the most commonly used, interpreters will choose other scales to enhance certain aspects of the seismic data.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
195
Fig. 4 Žcontinued..
information present in the troughs can be overlooked ŽBrown, 1996a.. With the advent of interactive computer displays, variable density displays have become the norm. In this display, the peaks and troughs are arbitrarily assigned distinct colours. One commonly employed colour scheme shows peaks as blue Žwith stronger positive values being shown as darker blue. passing through white Žnear zero amplitudes. to red troughs Žmore negative values as darker red.. Other colour schemes can be generated by interpreters to highlight specific aspects of the seismic data Že.g., bright spots, dim spots. on which they wish to focus. The variable density display gives the appearance of reflection continuity throughout the entire 3-D volume ŽFig. 4.. The 3-D seismic data volume is stored in digital format, on disk. The types of displays that can be generated ŽFig. 5. depend on the software and hardware capabilities of the interpreter, but generally can be grouped into a few distinct categories that are discussed next. 3.1. Vertical transects Vertical transects through a 3-D seismic volume ŽFig. 5a–d. look like 2-D seismic profiles, but differ in that their location and orientation are decided by
the interpreter in an interactive manner, rather than being constrained by the original seismic survey line orientation as is the case for 2-D data. Since the data are stored digitally, the interpreter can also zoom in on small portions of the seismic data, or zoom out to see the larger structural and stratigraphic framework. Most software packages allow the user to quickly display vertical transects: Ža. in the inline or line direction, Žb. in the crossline or trace direction, and Žc. arbitrary lines that represent transects through the data in any direction decided upon by the interpreter. The arbitrary line may consist of a single transect or a series of continuous transects that zigzag their way through the data set Žsometimes called a multipanel display.. Arbitrary lines are used when the interpreter wishes to view the true geometry of structural or stratigraphic features that are oriented obliquely to the line or trace orientation Žsee below., or when wishing to tie borehole log information from more than one well by integrating with the seismic data. With vertical transects, the data can be flattened on a selected horizon to more clearly view true stratigraphic relationships in areas that have been structurally deformed. This latter process is akin to using a formation top or log pick as a datum to construct a stratigraphic cross-section Žusing well logs or outcrops. rather than generating a structural cross-section.
196
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
Fig. 5. Different types of 3-D seismic displays that can be viewed during an interactive interpretation session. Ža, b. Line and trace transects that correspond to source and receiver line orientations, Žc. arbitrary transect, Žd. multipanel display, Že. time slice, Žf. horizon slice, Žg. perspective display, Žh. cube display. See text for discussion.
3.2. Horizontal sections These displays, known as time slices, represent a slice through the data at a given two-way traveltime ŽTWT, constant z coordinate; Fig. 5e and Fig. 6.. The display is somewhat analogous to a geologic map. The difference is that instead of viewing how stratigraphic units intersect the ground surface Žwhich may or may not be planar., the interpreter sees how the seismic manifestation of the stratigraphy intersects an arbitrarily selected plane of constant TWT through the seismic data. In both cases, however, the thickness of the reflection from a given stratigraphic unit on the display Žmap. is a function of the stratigraphic dip Žfor a constant thickness, a less steeply dipping bedrreflection will appear wider; Fig. 7a,b. and ‘‘thickness’’ Žfrequency. of the reflection event Žfor a given structural dip, higher frequency events will appear thinner; Fig. 7c.. Although Brown Ž1996a. and others recommended using time slices for horizon interpretation, most interpreters tend to concentrate their use of these displays on interpreting faults — especially where stratigraphic dips are small Ži.e., relatively
undeformed basins.. The utility of time slices for horizon interpretation is greatest when beds have a pronounced dip. In this case, horizon mapping on time slices can be a quick way of generating time structure maps for those horizons. With the advent of automatic horizon tracking and interfacing of seismic interpretation with mapping packages Žsee below., time structure maps can, in most circumstances, be generated just as readily by basing most interpretation on vertical sections. Fault interpretation and correlation can be significantly improved by interactively working with time slices and vertical slices together. Time slice displays show faults as curvilinear features that display lateral offsets of reflections, amplitudes or abrupt changes in seismic reflection character ŽFig. 6.. Typically, faults can be traced laterally for considerable distances on time slices, thus, reducing the ambiguity inherent in interpreting fault locations and orientations from 2-D seismic or well data Žsee below.. 3.3. Horizonr fault and map displays These displays show characteristics of horizons or faults that have been interpreted in a seismic volume.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
197
Fig. 6. Time slice through 3-D seismic volume, offshore Gulf of Mexico. Note curvilinear trends of reflection character offset that correspond to faults viewed in vertical sections. By interpreting the faults on both vertical transects and time slices, the potential for miscorrelation is reduced, leading to a more accurate interpretation than could be generated using 2-D seismic data.
They allow the interpreter to view spatial relationships in two dimensions. Time structure of an interpreted horizon, showing locations of faults, folds and structural dips, is perhaps the most commonly viewed display. If desired, the interpreter can interactively adjust the colour scale bar to detect structural relationships Že.g., areas with subtle closure that could act as hydrocarbon traps. that might otherwise be overlooked. In many areas, the seismic amplitudes associated with particular stratigraphic horizons can be of significance Že.g., Enachescu, 1993.. For example, ‘bright spots’ are associated with hydrocarbon accumulations in some areas, and interpreters will examine map displays of the amplitude of seismic horizons in order to look for stratigraphic or structural features that might be hydrocarbon traps. Displays of
horizon amplitudes are sometimes called horizon slices ŽFig. 5f.. Channel sandstones and other stratigraphic features can also, under some circumstances, be identified using seismic amplitudes Že.g., Hardage et al., 1994; Brown, 1996a.. Structural contours can be superimposed onto amplitude displays in order to allow the interpreter to easily search for relationships between these two data types ŽFig. 8.. Bouvier et al. Ž1989. present an example of the use of fault slices to examine fault sealing capabilities of Tertiary faults in the Niger Delta area. By generating fault slices in both the hanging wall and footwall, the juxtaposition of lithologies across the fault can be assessed. In this way, it might be possible to judge whether a fault is a barrier to fluid flow Že.g., sand on shale contact. or not Že.g., sand on sand. along the entire fault plane.
198
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
times employed in a ‘quick look’ fashion to identify stratigraphic configurations or rock properties elements that warrant detailed investigation. Horizon attributes Ždiscussed below. are also viewed in map views. 3.4. PerspectiÕe displays This type of display ŽFig. 5g. shows horizons, faults and well data as 3-D perspective Žisometric. views that can be rotated to help the interpreter assess spatial relationships in 3-D. Additionally, they can be employed to quality check interpretations — for example, to ensure that horizon or fault picks are physically plausible. With some interpretation packages, it may be possible to superimpose ‘attributes’ such as seismic amplitude, isochrons Žthickness, in time units., etc. on top of the 3-D surface to more easily evaluate, for example, possible relationships between amplitudes and structure. Illumination angles and opacity might also be adjusted to help detect subtle structural trends. 3.5. Cube displays
Fig. 7. Schematic diagram illustrating the effects of: Ža. stratigraphic dip, and Žb. reflection frequency on the thickness of a reflection viewed on a time slice. Adapted from Brown Ž1996a..
Although strictly not related to one horizon, map displays are also employed to view thickness, in time Žisochrons. or depth Žisopach. units, between two stratigraphic horizons. Other interÕal attributes, such as the maximum amplitude between two horizons, can also be viewed and interpreted in this fashion. These, and other interval attribute analyses are some-
This type of display ŽFig. 5h and Fig. 9. allows the interpreter to view the seismic data as a volume. By scrolling through the data cube Žfront to back, left to right, and top to bottom., the interpreter can quickly get an intuitive feel for the broad scale stratigraphic and structural configuration of a study area before beginning detailed interpretation. Interpretations Žfault and horizon. can, however, be made on the faces of the data cube. The data may be ‘clipped’ in various ways to visualize specific aspects of the data set that will assist in the interpretation. Most cube displays show the faces of the data volume, while data behind those faces remain out of
Fig. 8. Ža. Vertical transect through 3-D seismic volume from the San Juan Basin, New Mexico, showing classic ‘‘doublet’’ response Žtroughs. associated with thick accumulations of aeolian Jurassic Entrada Formation. Strong peak above the Entrada is the Todilto Formation Žgypsum and limestone.. Žb. Time structure contours of the Entrada horizon superimposed on horizon slice of the formation. Contours were generated by picking the horizon on seed lines, autotracking the horizon throughout the survey area, quality checking the result, then contouring the data. The contours suggest a thick build-up of the dune deposits along the lower margin of the survey area. The amplitude of the Todilto reflection decreases over the crest of the buildup Žlower amplitudes shown in lighter greys., helping to corroborate the interpretation and increase the attractiveness of this area as a drilling target Žsee Vincelette and Chittum, 1981, for a description of the seismic character of Entrada reservoirs.. Although small, these deposits can form attractive drilling targets. It would not be possible to pinpoint the optimal drilling location for these types of buildups, or other features such as pinnacle reefs, meandering channel sands, etc., with this degree of confidence using 2-D seismic or well control alone.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
sight. Voxel rendering technologies allow the interpreter to make specific ranges of amplitudes transparent, leaving only particular ranges of amplitudes visible. The objective is to facilitate the 3-D viewing and interpretation of subsurface features, such as
199
bright spots associated with hydrocarbon accumulations, that have specific amplitude characteristics. This type of display can be especially useful, for example, when planning a deviated wellbore so that it penetrates multiple pay zones that manifest them-
200
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
selves as bright spots at distinct stratigraphic levels. Virtual reality technologies are being explored that allow the interpreter to view the data in 3-D or even to ‘‘enter’’ the seismic volume ŽDorn, 1998.. However, application of these technologies is not widespread. 3.6. Combination displays Some software applications allow several different types of data and interpretations to be visualized together Že.g., Fig. 10.. For example, it may be desirable to examine well paths, interpreted horizons, and some seismic data together. As is the case with the other types of displays, the objective is to help the interpreter to visualize the geology, engineering and other types of data in 3-D. Another important component is that these displays help the interpreter to present hisrher results to others Že.g., co-workers, management, investors. who have not been active in the interpretation process.
4. Interpreting 3-D seismic data In the petroleum industry, most 3-D seismic interpretation is conducted by members of multidisciplinary teams that are composed of geologists, geophysicists and engineers ŽHart, 1997.. This is because experience has shown that the ambiguity inherent in the seismic method Ždue to limitations on vertical and lateral resolution and the non-uniqueness of the seismic response. can be reduced by integrating geologic and petroleum engineering data and concepts ŽFig. 11.. As with other subsurface studies, the principal objectives of a 3-D seismic survey are definition of subsurface stratigraphy, structure and rock physical properties. Most such data sets are collected from areas of existing production, so that some subsurface control Že.g., wireline logs, paleontological data, en-
201
gineering data. exists already. Having this information ‘up front’ allows the interpretation team members to better exploit the massive amounts of data available to them in the 3-D seismic volume. Typically, however, an interpretation is not viewed so much as a ‘final product’ as a ‘work in progress’ that is to be updated and revised as new data become available through drilling. The establishment of development drilling plans Žfinding infill and step out drilling sites. is a common application of 3-D seismic, although most interpreters will also use the data for exploration purposes Že.g., looking for other, as yet unproductive, stratigraphic intervals.. There is increasing use of 3-D seismic data, particularly by large companies, as a purely exploration tool. It should be noted that although the vertical resolution of seismic data Žgenerally, only features 10 m or more in thickness will be resolved. is not as great as wireline logs or core Žmillimeter scale in core, decimeter scale in logs., the continuous subsurface coverage in a 3-D survey provides much better spatial control for mapping. For example, in North America, typical development drilling density might be 16 wells per section Ž2.59 km2 ., yielding 16 ‘control points’ Žwells with wireline logs. in that area. A 3-D seismic survey over that same area might have 33.5 m = 33.5 m bins, yielding 2304 control points in the same area. The result is that the 3-D seismic survey has 144 times more control points for mapping than even the relatively tightly spaced wells ŽRay, 1995.. As such, mapping of stratigraphic and structural features using 3-D seismic Žprovided that they can be detected seismically. is much more accurate than mapping based on interpretation between even densely spaced wells or 2-D seismic data. 4.1. Stratigraphic interpretation As with a 2-D seismic interpretation, the stratigraphic interpretation of a 3-D seismic volume typi-
Fig. 9. Sample cube displays. Ža. Entire data set showing east–west and north–south vertical transects along the front and right side faces of the cube, respectively, and a time slice on the top face. Žb. Entire time range is shown, but the front face has been cut to show an east–west transect about 2r3 of the way into the volume, showing only a limited part of the survey area. Žc. Entire area of survey is shown, but the top face has been cut down about 3r4 of the way into the volume. By interactively scrolling through a data cube such as this, the interpreter can quickly assess changes in stratigraphic and structural style with location and depth.
202 B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218 Fig. 10. Combination display showing a 2-D seismic line Žleft, colour., a portion of a 3-D seismic cube Žback, grey tones., a voxel display of a 2nd 3-D volume Žgreens and yellows, right., two interpreted horizons from the seismic data Žblue, pink., some well logs Žblue, coming down from top. and the locations of the 3-D seismic data sets Žyellow and red grid at bottom., 2-D seismic lines Žblue, at bottom. and wells Žred dots at bottom.. By integrating different types of data and interpretations into single displays such as this, multidisciplinary teams can make better drilling and reservoir management decisions. Figure courtesy of Landmark Graphics.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
203
Fig. 11. Bar graph showing responses of 133 seismic interpreters when asked about the need to integrate geologic and engineering data and concepts into a 3-D seismic interpretation. Experience shows that it is only through integration of the technology with other data types that the true potential of 3-D seismic technology is realized. From Hart Ž1997..
cally begins by tying wireline logs Žgeologic data. to the seismic data using Õelocity surÕeys, Õertical seismic profiles andror synthetic seismograms Že.g., Sheriff and Geldart, 1995.. From these initial seed points, horizons are interpreted on a grid of vertical transects. The grid allows the user to ‘box in’ the picks, helping to ensure a consistent interpretation. Picking a horizon on every line in a 3-D volume can be a labour intensive task, and so most software packages have an autotracking application that can be employed to automate the process. The grid of manual horizon picks forms a network of seed points, from which the software will attempt to track a pick of similar phase and amplitude throughout the seismic volume. Under ideal conditions, such as a continuous reflection that has significantly higher amplitude than surrounding events, it might be possible to autotrack a horizon based on a single seed point. Conversely, in areas of high noise or geologic complexity it may not be possible to autotrack a horizon at all. In any case, the results of the autotracking need to be checked and manually edited where necessary. The finished horizon pick should be both geologically and geophysically defensible. Once time structure maps have been derived, the horizons are
converted to true depths Že.g., meters. using available velocity data. Once a stratigraphic framework has been established, such as by defining and mapping flooding surfaces, unconformities or other significant and definable horizons, detailed stratigraphic analyses Žseismic facies analysis, reflection character analysis. are conducted on those intervals that are judged to be of particular stratigraphic importance. Clinoforms, channels, parasequences, reef complexes or other stratigraphic entities are studied using conventional seismic stratigraphic criteria Že.g., Mitchum et al., 1977; Sarg, 1987. and by integrating all available log, core and biostratigraphic data Že.g., Hart et al., 1997.. Where line or trace orientations are oblique to stratigraphic trends, arbitrary lines through a seismic volume will be selected to examine the true longitudinal and cross-sectional geometries of clinoforms, channels and other features. With most software packages, it is possible to superimpose wireline logs directly over the seismic data in vertical transects in order to help merge geologic Žlogs. and geophysical Žseismic. data and concepts ŽFig. 12.. To do so, some time-depth conversion information is required. Depending on the
204
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
Fig. 12. Example of a seismic view with well log overlay. Log curve shows g-ray Žincreasing to the right. with true well position indicated by vertical white lines. Note good correspondence between lithology contrasts as indicated by logs Žlow g-ray, clean carbonates; higher g-ray, dolomitic sandstones and siltstones. and the location of prominent reflections. This type of display is used to help verify picks, both log and seismic. Seismic transect shows basinward progradation Žto left. of mixed siliciclasticrcarbonate Permian shelfrslope in Delaware Basin, SE New Mexico.
types of logs employed, these displays can help to guide correlations from well to well, identify facies associations or stratal surfaces, identify fluid contacts, etc. Time and horizon slices can help the interpreter to identify and map features such as fluvial channels, sinkholes, Že.g., Hardage et al., 1994; Brown, 1996a.,
deltaic lobes ŽHart et al., 1997. and even meteorite impact structures ŽIsaac and Stewart, 1993.. In the petroleum industry, knowledge of the distribution of these depositional features can help to identify or predict sedimentary facies distributions, and thus the location of reservoir quality rock or barriers or baffles to subsurface fluid flow that might compartmen-
Fig. 13. Structure map Žlower left. of a horizon suggests the presence of a small NW–SE trending graben Ždarker greys are structurally X lower areas.. A seismic transect in the inline direction ŽA–A . does not clearly show the subtle fault on the southwest. This might be the X orientation of a 2-D seismic line. By viewing an arbitrary line that is perpendicular to the fault orientation ŽB–B ., the fault geometry becomes clearer. Although the fault is readily apparent, it may not be clear from this view how the horizons correlate from one fault block X Y to another. As such, a multipanel display was chosen that goes around the fault pinchout to the north ŽC–C –C .. This latter display clearly X shows reflection continuity from one block to the other, thus, confirming the interpretation shown in B–B . Example from New York State Žsee Hart et al., 1996..
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
205
206
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
Fig. 14. Two different horizon attribute maps for the same seismic horizon. Ža. Dip map, showing variability in dip amount for the surface Ždarker s greater dip.. Žb. Azimuth map, showing which way the horizon is dipping. Light source is from the upper right ŽNE., so that surfaces dipping to the SW appear dark. These maps show linear trends that indicate the presence of subtle faults which were not evident when interpreting vertical transects through this data volume.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
talize reservoir. Typically, vertical sections, time structure maps, and horizon displays will be evaluated simultaneously to assess a given prospect. 4.2. Structural interpretation Structural and stratigraphic interpretations necessarily feed off one another. For example, it is not possible to calculate throw on a fault without being able to identify common horizons on either side of the structure. Conversely, to correlate a horizon from one side of a fault to the other, the interpreter needs to understand the fault geometry Žnormal, reverse,
207
etc... As with stratigraphic interpretation, vertical transects, time slices, and other displays will all be utilized together during the structural interpretation procedure. Faults can be interpreted on both time slices and vertical transects, and the results of interpretation on one display can be viewed and used to guide fault picking on the other. The ability to view arbitrary lines through the seismic volume can, as with stratigraphic interpretation, have a significant beneficial impact on structural interpretations ŽFig. 13.. As with a stratigraphic interpretation, the structural interpretation begins by identifying the large
Fig. 15. Continuity cube Žsimilar to a Coherency Cubee. showing subtle faulting indicated by curvilinear trends of low coherence Ždark. and poor data quality area. As with the horizon attribute maps, these cubes help interpreters to recognize subtle structural features that otherwise might be undetected in vertical sections and time slices. Stratigraphic features, such as channels and reefs, can also sometimes be delineated with continuityrcoherency cubes.
208
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
scale faults, then successively mapping finer details. Faults can generally be detected when the throw is greater than 1r4 the wavelength ŽSheriff and Geldart, 1995.. Subtle faults that are not easily recognized on vertical transects can sometimes be detected by generating and examining horizon attributes Že.g., Dalley et al., 1989; Hoetz and Watters, 1992; Hesthammer and Fossen, 1997a; Fig. 14.. These can include dip Žin msrm., azimuth Ži.e., the direction that the surface is pointing, ranging from 08 to 3608., edge detection Žemphasizing discontinuities. and other properties. Depending on the orientation and dip of the fault with respect to stratigraphic horizons, any particular one of these displays might help to detect subtle structures that might have a significant impact on subsurface fluid flow. Coherency Cubee processing Žand similar techniques. generates a seismic attribute that quantifies the similarity between a given seismic trace and its neighbours ŽBahorich and Farmer, 1995.. This numerical measure is somewhat analogous to the ‘reflection continuity’ that seismic interpreters have been employing qualitatively for many years. When the waveshape for a trace in a given time window is similar to that for adjacent traces, as might be expected when the stratigraphy is continuous across an area, the coherency attribute calculated at that position is high. When there are significant differences between traces, as might be expected where the stratigraphy is offset by faulting, the coherency is low. This type of data is derived from a 3-D seismic volume, and can be viewed in the same way; typi-
cally cube displays or time slices are considered to be most revealing ŽFig. 15.. This attribute can very precisely and quickly reveal the location of subtle faults or stratigraphic features that might be otherwise missed. 3-D seismic analyses can often result in structure maps that are significantly different from structure maps based on 2-D seismic andror well control. That these maps truly are more accurate than the original maps has been empirically Žand frequently. demonstrated by drilling results, and by integration with other data types Že.g., Badgett et al., 1994; Brown, 1996a; Weimer and Davis, 1996.. 4.3. Rock properties Definition of subsurface rock properties, such as porosity or fluid saturation, directly from reflection seismic data continues to be a field of significant interest. Seismic inversion is a technique that attempts to reconstruct the subsurface physical properties distribution, usually directly from the processed seismic traces Že.g., Lindseth, 1979.. Many inversion algorithms have been developed, but the inversion process itself provides non-unique solutions — sometimes many different stratigraphic successions could have produced a given seismic section. Accordingly, interpreters generally attempt to constrain the inversion process with physical property information Ždensity, velocity. derived from well control. Acoustic impedance volumes are generated for interpretation in this way.
Fig. 16. Empirical, quantitative derivation of rock properties from seismic attributes. Ža. Values of selected attributes are extracted from the bins that correspond to well locations, and crossplotted against well-derived measurements. In some cases, no correlation is present ŽAttribute 1.. In other cases, there might be a negative correlation between attributes and log properties ŽAttribute 2. or a positive correlation between the two ŽAttribute 3.. Linear and polynomial regression approaches might be tried. Žb. Relationships might be sought out between a reservoir property and several attributes. Here, the 3-D graph, which can be interactively rotated, allows the interpreter to view statistical relationships between three variables. The size of the data point is proportional to its closeness to the observer. This type of interactive data exploration can help the interpreter to seek out correlations between seismic attributes and log-based physical properties. Multiple regression techniques can be used to improve the correlation between seismic attributes and reservoir properties. Žc. Once an empirical, numerical relationship between seismic attributes and rock properties has been established, it is used to assign rock properties to other areas within the 3-D survey area that lack well control. Here, a 2nd order multiple regression based on three attributes Žnot the three shown in Ža.. has been employed to derive the thickness of the reservoir interval having a porosity greater than 12% Žcontours.. The underlying grey scale shows structure, with structurally high areas being the lightest. Structure generally shows a dip from upper right to lower left. The thickest porous section Ž) 20 m. is located beneath the structural culminations in this Jurassic shelf margin carbonate buildup Žcenter of map., although some porosity also appears to be present on the seaward flank Žbottom. — possibly as talus accumulations. See Hart and Balch Ž1998. for further discussion.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
209
210
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
Complex trace attributes, such as amplitude, instantaneous frequency, reflection strength and instantaneous phase and many others ŽTanner and Sheriff, 1977; Brown, 1996b. are currently being analyzed and exploited qualitatively in the hopes that they contain information about the physical properties of the rocks being imaged. Although the analysis of these attributes is not new, the degree of vigour with which 3-D seismic interpreters are currently deriving and exploiting them ŽHart, 1997. can be related to: Ža. the direct way in which large amounts of digital data can be analyzed, and Žb. the ability to link log-derived physical properties from individual wells to a specific traces in a spatially continuous 3-D seismic volume. Seismic amplitudes are the most readily imaged and qualitatively interpreted attribute ŽEnachescu, 1993., although other attributes are exploited in a qualitative way as well Že.g., Hardage et al., 1996., either individually or collectively. Nonuniqueness of response Že.g., seismic amplitudes can be affected by changes in porosity, bed thickness, reflector geometry, fluid content, processing and other variables. should be an important consideration when interpreting such data. Using a relatively new technique, growing numbers of interpreters are attempting to correlate seismic attributes empirically with reservoir physical properties measured by borehole logs ŽSchultz et al., 1994.. The complex trace attributes potentially contain information about the physical properties of the rocks being imaged seismically, but the direct relationship between the rock properties and seismic attributes may be practically impossible to derive from first principles. The objective is to try to correlate physical properties, as measured from borehole logs, with seismic attributes derived from the traces that correspond to the boreholes. Different methods are being utilized or developed, including multiple regression, geostatistics, and neural networks first to derive the relationships, and then to distribute properties throughout the area of the seismic survey ŽFig. 16.. Matteucci Ž1996. and others have identified several aspects that need to be considered when adopting this approach. First and foremost, the correlation between the attributes and physical properties must be statistically significant. A prediction of reservoir properties based on a quantitative, empirical relation-
ship is of limited value if the correlation coefficient is too low. Second, the relationship should be statistically robust and take into account uncertainties, for example, in the exact borehole location or seismic noise. The location of boreholes might be known only to within a range of bins, especially if the wells are old Žand poorly surveyed. or large offsets in deviated wells are present. Using several adjacent bins in the correlation exercise can, provided that the reservoir is homogeneous enough, help to determine the robustness of the correlation. Finally, the number of attributes that can be generated is great, increasing the probability that there will be at least one statistically significant, but possibly spurious, correlation ŽKalkomey, 1997.. Additionally, Hirsche et al. Ž1997. pointed out that limited well control or biased sampling Žonly specific facies might be drilled. can make it difficult to assess whether basic statistical assumptions that underpin the utilization of these methods are violated. Ideally, the interpretation team might have a large number of wells to work with. From these wells, they can randomly exclude some from the calibration process, then test the predictions of their work against the data from the excluded wells Že.g., Schultz et al., 1994.. This approach may not be feasible in areas of limited well control Že.g., new or small fields.. One of the fundamental questions that should be asked of any rock properties prediction is whether the result makes geologic sense. This is true regardless of how the prediction is derived, or the strength of the statistical correlation. The most prudent approach to assessing these predictions is to view them as working hypotheses that: Ža. will help to assess the risk associated with drilling a specific prospect, and Žb. will need to be revised as new information becomes available. 4.4. Integration As is the case with any subsurface data, there is a certain amount of ambiguity in the interpretation of seismic data Ž2-D or 3-D. that cannot be removed by working with it alone. These limitations are the product of many factors, including the vertical resolution and non-unique response of seismic data, acquisition and processing programs, and interpreters’
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
skills or biases. As such, most interpreters recognize that their seismic interpretations need to be integrated with geologic and engineering data and concepts in order to maximize the benefit that they derive from a 3-D seismic volume ŽHart, 1997.. As noted previously, the integrated study might include paleontologic, core, wireline log, production, and other data types. Case studies presented by Badgett et al. Ž1994., Hardage et al. Ž1994., Rafalowski et al. Ž1996. and Hart et al. Ž1997. illustrate how 3-D seismic, core, wireline log, pressure and production data are integrated with sequence stratigraphic, seismic attribute, production and other analyses to generate holistic reservoir models. Detailed analyses of the seismic data themselves, possibly including wavelet extraction, forward modeling, phase correction and other steps, can be employed to add confidence to the ties between seismic and well data Že.g., Dorn et al., 1996.. Ideally, it would be possible to export a 3-D reservoir model populated with physical properties directly from the seismic analyses into a reservoir simulator, but this type of activity is not yet commonplace ŽHart, 1997.. Integration is a key component in the developing field of time-lapse or 4-D seismic analyses. This type of analysis seeks to find differences in amplitude Žor other attributes. from a given reservoir that might be visible in successive 3-D seismic surveys covering the same area, and then to relate these differences to changes in pore fluids that are due to production or enhanced recovery operations Že.g., Greaves and Fulp, 1987; Anderson et al., 1996.. This approach is most widely used where amplitudes Žbright spots. provide direct hydrocarbon indicators, although applications of the technique to monitor steamfloods or CO 2 injection programs have also been undertaken ŽD. Lumley, personal communication, 1998..
5. Discussion First conceived within large petroleum companies in the early 1960s, it was not until the 1970s that 3-D seismic concepts were publicly presented ŽWalton, 1972; French, 1974.. Documentation of 3-D seismic success stories in the late 1970s Že.g., West, 1979. to
211
middle 1980s fueled interest in the technology. Although much, if not most, of the ensuing drilling success attributable to 3-D seismic has gone unreported in the scientific literature Žthere are many case histories in trade journals., the worth of the technology is empirically demonstrated by a single statistic that shows the extent to which the petroleum industry has begun to focus on the technology: by the end of 1996, 3-D surveys accounted for nearly 75% of all seismic acquisition ŽSociety of Exploration Geophysics, 1997.. The application and development of this technology has had a tremendous impact on the petroleum industry, largely due to the precision and accuracy with which stratigraphic and structural elements can be identified, mapped and drilled Že.g., Haldorsen and Damsleth, 1993; Sibley and Mastoris, 1994.. These same types of benefits should be realizable in any geoscience discipline where the objective is to characterize properly subsurface structure, stratigraphy and sediment Žor rock. properties. 5.1. Non-petroleum applications The cost of acquiring and processing 3-D seismic data is no doubt the greatest impediment to the development of more widespread application of the technology. Additionally, relatively inexpensive PCbased data visualization and interpretation packages are making interpretation capabilities Žat least the more commonly used applications. accessible to greater numbers of potential users. Eaton et al. Ž1997. performed cost–benefit analyses to examine when the technology can be cost-effective in the mining industry, and also examined some of the technical considerations related to seismic imaging in plutonic environments, rather than layered sedimentary successions. For example, they noted that while most reflection seismic processing is geared towards enhancing continuous features such as bedding, in crystalline terranes scattering effects from localized bodies are of paramount importance. Unmigrated 3-D seismic volumes can help interpreters to recognize scattering anomalies produced by features such as ore bodies, and may therefore be of greater interest in the mining industry than migrated volumes ŽEaton et al., 1997.. Cost was initially considered to be an obstacle to the application of 3-D seismic in the environmental industry ŽHouse et al., 1996., although recent experi-
212
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
mental studies are suggesting otherwise ŽSiahkoohi and West, 1998.. Cost–savings realized by technological improvements Že.g., more cost-efficient acquisition layouts. are likely to overcome this barrier, especially in hydrogeology where aquifers are structurally or stratigraphically complex, and the economic considerations associated with water availability or contamination are great. Problems associated with obtaining high quality shallow seismic data Žespecially in the vadose zone. are also currently an impediment. As an ‘academic’ discipline, structural geology could benefit by more widespread application of 3-D seismic data. Experience in the petroleum industry has shown that many structural interpretations based on 2-D seismic are wrong, to varying degrees Žtypically, fault geometries are oversimplified; e.g., Brown, 1996a; Mansfield, 1996.. Since 2-D seismic based interpretations are used as input into kinematic models, it may be desirable or even necessary to revise some of the models to take into account new details that could be provided by 3-D seismic mapping. In some cases, it will not be possible to derive the true 3-D geometry Žand evolution. of these structures without 3-D seismic data. For example, Mansfield Ž1996. and Rowan et al. Ž1998. demonstrated how 3-D seismic data can be employed to understand the spatial and temporal development of a growth fault arrays. These types of analyses traditionally have been based on 2-D analyses Žseismic or outcrop., either in section or plan, and case studies that document the true 3-D geometry and evolution of real fault arrays are rare. Beginning in the 1970s, the development of seismic stratigraphy Žand the offshoot of sequence stratigraphy. as a mature science did much to revolutionize and reinvigorate the field of stratigraphy. The first order controls on depositional sequence architecture Žsea-level change, subsidence and sediment supply. are fairly well established, although the relative importance of each and the nomenclature used to describe successions of sedimentary rocks remain topics of considerable debate. By providing a continuous and accurate image of the subsurface, 3-D seismic data have the potential to improve our understanding of some of the 2nd or higher order controls on stratigraphic architecture. Although 3-D seismic data are often collected over smaller areas than 2-D
seismic surveys, limiting their usefulness for regional stratigraphic interpretations, the lateral continuity provided by 3-D seismic will help stratigraphers to better understand the controls of local processes such as reactivation of basement tectonic elements, and autocyclic lobe switching on depositional architecture. Additionally, many studies have shown that stratal geometries derived from 2-D seismic profiles can be misleading, possibly leading to erroneous reconstructions of depositional histories Že.g., Hart and Long, 1996.. The integration of 3-D seismic data into sequence stratigraphic studies will help to reduce these potential errors. 5.2. Limitations of 3-D seismic technology Experience in the petroleum industry has shown that, despite its general applicability, 3-D seismic technology Žlike any technology. has its limitations. These limitations can influence how a survey is collected, processed and interpreted. There are areas where good quality seismic data Ž2-D or 3-D. are impossible or very difficult to obtain. For example, velocity-related problems have, until recently, prevented geophysicists from obtaining clear subsurface images below salt bodies in the Gulf of Mexico or beneath thrust sheets. New processing flows, for example, pre-stack depth migration, are helping in some of these areas, however, there remain areas where seismic acquisition is not Ž?yet. feasible. Portions of North America’s Permian Basin region fall into this category. In some areas of this basin, unconsolidated Tertiary sediments overlie and bury a dissolution surface developed on a Permian aged salt zone. The thickness of the Tertiary fill changes abruptly and there is a strong velocity contrast between the fill and the underlying Paleozoic rocks. For this reason, much energy is scattered at this contact and the implementation of certain key processing steps becomes problematic. As such, it is not possible to obtain useable images of much of the Paleozoic section. In shallow earth applications, it is generally difficult to obtain good quality seismic data from unconsolidated or poorly consolidated soils that are above the water table. These, and some other settings, are considered to be ‘‘lossy’’ media Ži.e., much energy is lost through attenuation as the acoustic pulse
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
propagates. where it may be difficult or impossible to obtain interpretable seismic images.
213
In other areas, noise generated by waves, wind, or ‘‘cultural’’ features Že.g., working pumpjacks, road
Fig. 17. Example showing the effects of seismic processing on data interpretability. The images show a common vertical transect through two different versions of a 3-D seismic volume. Both data versions began with the same field data input, but different processing flows were applied Ždiscussion and analysis of the processing differences is beyond the scope of this paper.. The upper Žoriginal. image is difficult to interpret above 600 ms ŽTWT., but after reprocessing Žlower image. the stratigraphy and structure in that part of the section become clearer. Conversely, the reprocessed volume Žlower. has some of the higher frequencies removed, making stratigraphic details somewhat harder to interpret in the lower part of the section.
214
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
traffic, electromagnetic emissions from power cables. can cause low signal-to-noise ratios that make data difficult to interpret. Source testing, wherein different types of source are tested to determine which provides the best combination of frequency and energy characteristics for a particular application, and noise testing, looking at ambient and system noise levels, can be used to help design survey acquisition parameters or even determine whether it is possible to collect usable seismic data. Other concerns are related to acquisition parameters Žsuch as survey design., processing parameters, the subsurface geology, non-uniqueness of the seismic response and, perhaps, the way that these elements interact. For example, the acquisition footprint is an artifact characterized by patterns of noise in 3-D data that are related to the geometric distribution of sources and receivers on the earths surface ŽMarfurt et al., 1998.. This, and noise related to other sources Že.g., multiples, diffractions, low fold areas., can obscure stratigraphic and structural features Že.g., Hesthammer and Fossen, 1997b. and disrupt seismic attribute studies Že.g., Marfurt et al., 1998.. As illustrated in Fig. 17, choices made during the processing phase can significantly impact seismic data interpretability. Older seismic data sets Žboth 2-D and 3-D. are sometimes reprocessed using different parameters to enhance interpretability Že.g., Hesthammer and Fossen, 1997a.. Processing can affect the quality of the final data volume for a number of reasons. Two of the more important aspects include the experience of the processors in working with data from a particular area Že.g., there might be velocity problems or multiples that a processor with experience from that area might know how to handle., and the choice of processing algorithms Žoften the better algorithms take more computer time and are therefore more costly.. Assuming that the data set has been satisfactorily acquired and processed, there are still aspects of the interpretation phase that can affect the validity of the results. The most obvious of these are the interpreters capabilities and experience working with 3-D seismic data and the software used in the interpretation. Various types of pitfalls can confront even an experienced interpreter. Velocity related problems Že.g., a horizontal layer might appear to be ‘‘pulled
up’’ on a seismic transect beneath a region that has fast velocities. are a common type of pitfall, and can be most problematic when available velocity control is limited and lateral lithologic changes and depths are great. Depth migration converts the z axis of the seismic volume from time to depth Žeither pre- or post-stack., but the quality of the result depends on the accuracy of the velocity model used during the migration. Experience, and an understanding of the principles of geology and geophysics can help the interpreter to watch out for these and other types of interpretation problems. Finally, within the petroleum industry there is general recognition of the limitations of 3-D seismic technology, and these limitations need to be remembered if, or when, the technology is applied in other fields. For example, in addition to the artifacts and pitfalls just described, practical limits on vertical and lateral resolution can make targets impossible to identify in seismic data. Improper survey design and excessive cost cutting during acquisition and processing have been known to lead to poor data quality and failure in the petroleum industry. The geology of an area and the project economics need to be characterized up front, in order to properly design a 3-D survey or, indeed, to decide whether a 3-D seismic survey is worth acquiring. Likewise, although 3-D seismic data can provide some valuable insights compared to 2-D seismic or well data, it is only through the integration of multidisciplinary data sets and concepts Že.g., pressures, borehole logs, core, geochemical analyses. that the full benefits of 3-D seismic technology are realized. No matter how widespread the technology has become in the petroleum industry, economic, technical or other obstacles can cause 3-D seismic to be an impractical tool for studying the subsurface. Overcoming these obstacles will continue to be a topic of considerable interest. 6. Conclusions This summary has endeavoured to illustrate how 3-D seismic technology is applied in the petroleum industry and to suggest that similar approaches might be undertaken in other branches of the earth sciences. Although a 3-D seismic data volume provides a tremendous amount of information about the sub-
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
surface structure, stratigraphy and rock properties, it is only through integration of these geophysical data with other geologic and engineering data and concepts that the true potential of the technology is unlocked. The quality and continuity of the subsurface image obtained with a properly acquired 3-D survey, combined with the flexible data visualization and interpretation capabilities of interpretation software and data integration, help petroleum geoscientists to produce the world’s hydrocarbon resources more economically and efficiently. These, or similar benefits may be realizable in other applied and fundamental branches of the geosciences.
Acknowledgements The material presented in this paper is derived from the author’s own published and unpublished research, and from examples and concepts presented in the literature. Geographic coordinates for some of the illustrations have been deliberately disguised or omitted to conserve the confidentiality of the data. Support for the author’s current research comes from the AdÕanced ReserÕoir Management Project of Los Alamos National Laboratory, the Southwest Section of the Petroleum Technology Transfer Council, and the New Mexico Bureau of Mines and Mineral Resources. Data sets and technical guidance have been provided by Amoco, Ardent Resources, Cross Timbers Oil, Harvey E. Yates, Marathon, and Smacko Operating. Interpretation software has been provided to the New Mexico Institute of Mining and Technology by Landmark Graphics Corporation.
Appendix A. Glossary of some commonly used terms in 3-D seismic interpretation This Appendix provides definitions for some terms in 3-D seismic technology based on common usage by interpreters. Some of these terms have other meanings in the field of reflection seismology. Arbitrary line r transect — A vertical seismic transect through a 3-D seismic volume, the location, orientation and length of which is defined freely by the interpreter.
215
Autotracking — Automated picking of a seismic horizon throughout a 3-D seismic volume based on seed points provided by the interpreter. Picking parameters are set by the interpreter. This process is used to provide a continuous horizon pick more rapidly than could be achieved through manual picking. Despite the time efficiency, results of autotracking applications need to be quality checked by interpreters, especially in areas of complex geology or poor data quality. Bin — A rectangular area in the x, y plane Žhorizontal. that is represented by a single seismic trace in a 3-D volume. Bin dimensions are determined by acquisition parameters, and each bin can be identified by a specific combination of line and trace numbers. Coherency attribute — An attribute defined by comparing a seismic trace with its neighbours over a sliding time window in order to numerically define the lateral trace continuity at that location. This attribute is commonly used to define discontinuities that can be due to faulting or stratigraphic features Že.g., channel or reef margins.. The term Coherency Cubee refers to a technology that has been patented by Amoco, but other companies have developed similar attributes Žmost of which are proprietary.. Complex trace attribute — A seismic attribute Že.g., instantaneous phase, instantaneous frequency. that is derived from the Hilbert transform. Crossline (trace) — One of the two sequentially numbered orthogonal reference coordinates for 3-D seismic surveys Žthe other being the linerinline.. For land surveys, the crossline orientation is typically the direction in which source lines were laid out. Fault slice — Amplitudes extracted along the plane of a fault, and viewed projected onto a vertical plane. Fold (multiplicity) — The number of field traces sharing a common midpoint that are stacked together during data processing to produce a single trace in the final data volume. All else being equal, the higher the fold, the better the data quality. The fold may not be constant throughout the entire area of a 3-D survey. Horizon — A seismic pick that typically corresponds to a particular stratigraphic level Že.g., the top of a formation..
216
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
Horizon attributes — Attributes that are derived from and define geometric properties of a picked, continuous horizon, such as dip, azimuth, edge detection, etc. Horizon slice — Instantaneous amplitudes extracted along a horizon and displayed in a map view. Inline (line) — One of the two sequentially numbered orthogonal reference coordinates for 3-D seismic surveys Žthe other being the tracercrossline.. For land surveys, the inline orientation is typically the direction in which receiver cables were laid out. InterÕal attribute — A seismic attribute Že.g., average frequency, average amplitude. that is derived for an interval that is defined either by two horizons or a user selected time window defined with respect to one horizon. Migration — A seismic processing step which is used to reposition reflected energy to its true subsurface location Žthus, eliminating sideswipe., collapse diffractions and shrink the Fresnel zone. 3-D migration produces a seismic image that is more accurate than an equivalent 2-D image. Multipanel display — A display that consists of two or more contiguous vertical transects Žtypically arbitrary lines. through a 3-D seismic volume. Patch — In land 3-D seismic acquisition, a rectangular series of geophones that record reflections from a shotpoint that is typically located near the center of the patch. PerspectiÕe display — A display that shows objects from a seismic survey, typically interpreted horizons and faults, in pseudo 3-D form. These displays can be rotated Ži.e., viewed from any angle. and are useful visualization tools for quality checking interpretations or presenting results. Sideswipe — Reflected energy that comes from objects to the side of the vertical plane that is intended to be imaged in a 2-D seismic transect. Sideswipe is eliminated during 3-D migration. Synthetic seismogram — Seismic traces generated by mathematically convolving a seismic wavelet with a time series of reflection coefficients that has been derived from well log data Žvelocity and density logs.. These ‘‘synthetics’’ are used to predict the seismic response at a borehole location and thus tie log-derived geologic information to the seismic data. Time lapse (4-D) seismic — The use of different vintages of 3-D seismic data covering a constant
area to monitor changes in subsurface conditions. A common example is the use of changes in seismic amplitudes through time to monitor changes in the location of an oilrwater contact that are due to production. Timeslice — A slice through a 3-D seismic volume at a constant TWT. Since subsurface velocities may vary laterally, a constant TWT does not necessarily represent a constant subsurface depth. Velocity surÕey — Determination of the subsurface velocity field, usually by lowering a receiver down a borehole to different depths, generating an acoustic pulse at the surface, then measuring the vertical traveltime to the receiver. Vertical seismic profile (VSP) — A seismic profile that is collected in a manner somewhat analogous to a velocity survey, but with more recording depths. Longer recording times are analyzed in order to detect reflected energy from subsurface interfaces, and not just first arrivals. Processing transforms the raw data into seismic traces that: Ža. show the seismic response in the vicinity of the borehole, Žb. can be directly tied to other borehole information Že.g., wireline logs., and Žc. can be displayed in either time or depth. Voxel — A ‘‘ volume element’’ defined by the bin size Žin the x, y plane. and the sampling interval Žin the z direction.. For conventional 3-D seismic volumes, each voxel stores a numerical amplitude value. References Anderson, R.N., Boulanger, A., He, W., Sun, Y.F., Xu, L., Hart, B.S., 1996. 4-D seismic monitoring of drainage in the Eugene Island 330 Field in the Offshore Gulf of Mexico. In: Weimer, P., Davis, T. ŽEds.., Applications of 3-D Seismic Data to Exploration and Production, Vol. 42. AAPG Studies in Geology Series, pp. 9–19. Arestad, J.F., Davis, T.L., Benson, R.D., 1996. Utilizing 3-D, 3-C seismology for reservoir property characterization at Joffre Field, Alberta, Canada. In: Weimer, P., Davis, T. ŽEds.., Applications of 3-D Seismic Data to Exploration and Production, Vol. 42. AAPG Studies in Geology Series, pp. 171–178. Badgett, K.L., Hill, P.L., Mills, W.H., Mitchell, S.P., Vinson, G.S. III, Wilkins, K.L., 1994. Team combines technologies to target horizontal wells in Gulf of Mexico Oil Field. Oil Gas J. 14, 44–49. Bahorich, M., Farmer, S., 1995. 3-D seismic discontinuity for faults and stratigraphic features: the coherence cube. The Leading Edge 14, 1053–1058.
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218 Bouvier, J.D., Kaars-Sijpesteijn, C.H., Kluesner, D.F., Onyejekwe, C.C., Van der Pal, R.C., 1989. Three-dimensional seismic interpretation and fault sealing investigations, Nun River Field, Nigeria. Am. Assoc. Pet. Geol. Bull. 73, 1397– 1414. Brown, A.R., 1996a. Interpretation of 3-dimensional seismic data, 4th edn. American Association of Petroleum Geologists Memoir 42, 436 pp. Brown, A.R., 1996b. Seismic attributes and their classification. The Leading Edge 15, 1090. Crawley Stewart, C.L., 1995. 3-D solution to a 2-D pitfall: seismic detection of carbonate buildups, Kiva Field, Paradox Basin, San Juan County, Utah. In: Ray, R.R. ŽEd.., High-definition Seismic: 2-D, 2-D Swath, and 3-D Case Histories. Rocky Mountain Association of Geologists, Denver, pp. 177– 183. Cross, T.A., Lessenger, M.A., 1988. Seismic stratigraphy. Annu. Rev. Earth Planet. Sci. 16, 319–354. Dalley, R.M., Gevers, E.C.A., Stampfli, G.M., Davies, D.J., Gastaldi, C.N., Ruijtenberg, P.A., Vermeer, G.J.O., 1989. Dip and azimuth displays for 3-D seismic interpretation. First Break 7, 86–95. Dorn, G.A., 1998. Modern 3-D seismic interpretation. The Leading Edge 17, 1262–1272. Dorn, G.A., Tubman, K.M., Cooke, D., O’Connor, R., 1996. Geophysical reservoir characterization of Pickerill Field, North Sea, using 3-D seismic and well data. In: Weimer, P., Davis, T. ŽEds.., Applications of 3-D Seismic Data to Exploration and Production, Vol. 42. AAPG Studies in Geology Series, pp. 107–122. Eaton, D.W., Milkereit, B., Adam, E., 1997. 3-D seismic exploration. In: Gubins, A.G. ŽEd.., Proceedings of Exploration 97: Fourth Decennial International Conference on Mineral Exploration, pp. 65–78. Enachescu, M.E., 1993. Amplitude interpretation of 3-D reflection data. The Leading Edge 12, 678–685. French, W.S., 1974. Two-dimensional and three-dimensional migration of model-experiment reflection profiles. Geophysics 39, 265–277. Greaves, R.J., Fulp, T.J., 1987. Three-dimensional seismic monitoring of an enhanced oil recovery process. Geophysics 52, 1175–1187. Haldorsen, H.H., Damsleth, E., 1993. Challenges in reservoir characterization. Am. Assoc. Pet. Geol. Bull. 77, 541–551. Hardage, B.A., Levey, Pendleton, V., Simmons, J.L. Jr., Edson, 1994. 3-D seismic case history evaluating fluvially deposited thin-bed reservoirs in a gas-producing property. Geophysics 59, 1650–1665. Hardage, B.A., Carr, D.L., Lancaster, D.E., Simmons, J.L. Jr., Hamilton, D.S., Elphick, R.Y., Oliver, K.L., Johns, R.A., 1996. 3-D seismic imaging and seismic attribute analysis of genetic sequences deposited in low-accommodation conditions. Geophysics 61, 1351–1362. Hart, B.S., 1997. 3-D seismic: what makes interpreters tick?. The Leading Edge 16, 1114–1119. Hart, B.S., Balch, R.S., 1998. Approaches to defining reservoir physical properties from 3-D seismic attributes with limited
217
well control: an example from the Jurassic Smackover Formation, Alabama. In: DeMis, W.D., Nelis, M.K. ŽEds.., The Search Continues into the 21st Century, Vol. 98–105. West Texas Geological Society Publ., pp. 207–212. Hart, B.S., Long, 1996. Forced regressions and lowstand deltas: Holocene Canadian examples. J. Sediment. Res. 66, 820–829. Hart, B.S., Copley, D., Loewenstein, S., 1996. Forging partnerships: chasing the Rose Run Play with 3-D seismic in the Empire State. Oil Gas J. 94 Ž42., 88–91. Hart, B.S., Sibley, D.M., Flemings, P.B., 1997. Seismic stratigraphy, facies architecture and reservoir character of a Pleistocene shelf margin delta complex, Eugene Island Block 330 Field, Louisiana Offshore. Am. Assoc. Pet. Geol. Bull. 81, 380–397. Hesthammer, J., Fossen, H., 1997a. Seismic attribute analysis in structural interpretation of the Gullfaks Field, northern North Sea. Pet. Geosci. 3, 13–26. Hesthammer, J., Fossen, H., 1997b. The influence of seismic noise in structural interpretation of seismic attribute maps. First Break 15, 209–219. Hirsche, K., Porter-Hirsche, J., Mewhort, L., Davis, R., 1997. The use and abuse of geostatistics. The Leading Edge 16, 253–260. Hoetz, H.L.J.G., Watters, D.G., 1992. Seismic horizon attribute mapping for the Annerveen Gasfield, the Netherlands. First Break 10, 41–51. House, J.R., Boyd, T.M., Haeni, F.P., 1996. Haddam Meadows, Connecticut: a case study for the acquisition, processing, and relevance of 3-D seismic data as applied to the remediation of DNAPL contamination. In: Weimer, P., Davis, T. ŽEds.., Applications of 3-D Seismic Data to Exploration and Production, Vol. 42. AAPG Studies in Geology Series, pp. 257–266. Isaac, J.H., Stewart, R.R., 1993. 3-D seismic expression of a cryptoexplosion structure. Can. J. Explor. Geophys. 29, 429– 439. Jackson, W., Hilterman, F., 1979. Use of seismic modeling for reef exploration in Michigan Basin. Geophysical Development, unpublished report. Kalkomey, C.T., 1997. Potential risks when using seismic attributes as predictors of reservoir properties. The Leading Edge, pp. 247–251. Kanasewich, E.R. et al., 1987. Seismic studies of the crust under the Williston Basin. Can. J. Earth Sci. 24, 2160–2171. Kanasewich, E.R., Burianyk, M.J.A., Dubuc, G.P., Lemieux, J.F., Kalantzis, F., 1995. Three-dimensional seismic reflection studies of the Alberta basement. Can. J. Explor. Geophys. 31, 1–10. Lindseth, R.O., 1979. Synthetic sonic logs — a process for stratigraphic interpretation. Geophysics 44, 3–26. Mansfield, C.S., 1996. Fault displacement distributions: insights using high-resolution 3-D seismic data. In: Weimer, P., Davis, T. ŽEds.., Applications of 3-D Seismic Data to Exploration and Production, Vol. 42. AAPG Studies in Geology Series, pp. 231–236. Marfurt, K.J., Scheet, R.M., Sharp, J.A., Harper, M.G., 1998. Suppression of the acquisition footprint for seismic sequence attribute mapping. Geophysics 63, 1024–1035. Matteucci, G., 1996. Seismic attribute analysis and calibration: a
218
B.S. Hart r Earth-Science ReÕiews 47 (1999) 189–218
general procedure and a case study. SEG Annual Meeting, Expanded Abstracts, pp. 373–376. Mitchum, R.M., Vail, P.R., Thompson, S., 1977. The depositional sequence as a basic unit for stratigraphic analysis. In: Payton, C.E. ŽEd.., Seismic Stratigraphy — Applications to Hydrocarbon Exploration. AAPG Memoir 26, pp. 53–62. Rafalowski, J.W., Regel, B.W., Jordan, D.L., Lucidi, D.O., 1996. Green Canyon Block 205: lithofacies, seismic facies and reservoir architecture, In: Weimer, P., Davis, T. ŽEds.., Applications of 3-D Seismic Data to Exploration and Production, Vol. 42. AAPG Studies in Geology Series, pp. 133–142. Ray, R., 1995. Utilizing seismic for higher definition of geologic prospects: 2-D, 2-D swath, and 3-D seismic. In: Ray, R.R. ŽEd.., High-definition Seismic: 2-D, 2-D Swath, and 3-D Case Histories. Rocky Mountain Association of Geologists, Denver, pp. 1–6. Rowan, M.G., Hart, B.S., Nelson, S., Flemings, P.B., Trudgill, B.D., 1998. Three-dimensional geometry and evolution of a salt-related growth-fault array: EI 330 Field, offshore Louisiana, Gulf of Mexico. Mar. Pet. Geol. 15, 309–328. Sarg, J.F., 1987. Carbonate sequence stratigraphy. In: Wilgus, C.K., et al. ŽEds.., Sea-level Changes — An Integrated Approach, Vol. 42. SEPM Special Publication, pp. 155–181. Schultz, P.S., Ronen, S., Hattori, M., Corbett, C., 1994. Seismicguided estimation of log properties: Part 1. A data-driven interpretation methodology. The Leading Edge, pp. 305–315. Sheriff, R.L., Geldart, 1995. Exploration Seismology, 2nd edn. Cambridge Univ. Press, New York, 592 pp. Shipley, T.H., Moore, G.F., Bangs, N.L., Moore, J.C., Scoffa,
P.L., 1994. Seismically inferred dilatancy distribution, northern Barbados Ridge decollement: implications for fluid migration and fault strength. Geology 22, 411–414. Siahkoohi, H.R., West, G.F., 1998. 3-D seismic imaging of complex structures in glacial deposits. Geophysics 63, 1041– 1052. Sibley, D.M., Mastoris, S.S., 1994. Rethinking conventional field development: 3-D shows direct detection of bypassed oil reserves. In: Society of Petroleum Engineers International Petroleum Conference and Exhibition of Mexico Proceedings, SPE Paper 28719, pp. 509–519. Society of Exploration Geophysics, 1997. SEG seismic activity survey. The Leading Edge 16, 77–78. Tanner, M.T., Sheriff, R.E., 1977. Application of amplitude frequency and other attributes to stratigraphic and hydrocarbon determination. In: Payton, C.E. ŽEd.., Seismic Stratigraphy — Applications to Hydrocarbon Exploration. American Association of Petroleum Geologists Memoir 26, Tulsa, pp. 301–327. Vincelette, R.R., Chittum, W.E., 1981. Exploration for oil accumulations in Entrada Sandstone, San Juan Basin, New Mexico. Am. Assoc. Pet. Geol. Bull. 65, 2546–2570. Walton, G.G., 1972. Three-dimensional seismic method. Geophysics 37, 417–430. Weimer, P., Davis, T., 1996. Applications of 3-D seismic data to exploration and production, Vol. 42. AAPG Studies in Geology Series, 270 pp. West, J., 1979. Development near for Thailand field. Oil Gas J. 77 Ž32., 74–76.