Enhanced velocity analysis, binning, gap infill, and imaging of sparse 2D/3D seismic data byCRS techniquesGuido Gierse*, Henning Trappe, Juergen Pruessmann, Gerald Eisenberg-Klein, TEECJeremy Lynch, Perenco, Doug Clark, Chevron

Summary

CRS processing does not only provide enhanced images ofsparse 2D and 3D seismic data, but also adds to animproved preprocessing. This is demonstrated by severaldata cases. A pseudo-3D imaging was tested on crookedline 2D data showing an areal extension with a low-fold 3Dcoverage. 3D CRS processing can provide a goodsubsurface stack while preserving the fine low-fold binninggrid, whereas conventional imaging would increase thesubsurface fold by coarsening the bin size. Anotherproblem of sparse data are data gaps due to surface mutesor missing CMP traces. CRS processing automaticallycloses these gaps by a dip-consistent interpolation andextrapolation of seismic data. Finally, stacking velocityanalysis in sparse 3D data is most often ambiguous due tomissing offsets and low signal-to-noise ratio in the prestackdata. Partial CRS stacking and regularization of theprestack data improve the data quality of so-called CRSgathers that allow a reliable velocity-dependent semblance,moveout, and stack power analysis while preserving themoveout in the data.

Introduction

Exploration projects frequently have to rely on sparse 2Dand 3D seismic data, either due to the incorporation of oldlow-fold data, or due to limited acquisition layouts in recentsurveys. Seismic surveying is often constrained in time byseasonal or contractual restrictions of the access to thesurvey area. In addition, technical reasons can keep partialareas out of reach, e.g. extremely rugged terrain, swamp ortransition zones. Finally, limited budgets frequentlyoverride all technical and organizational capabilities.

In each of these cases, the prospect evaluation is partlybased on sparse seismic data that may exhibit irregularacquisition geometries and low or zero CMP fold in someregions of the survey. Data gaps in initial stack sections aredue to strongly varying surface mutes at irregulardistributions of near-offset traces and to missing traces atzero-fold locations. On the whole, the sparse seismicacquisition implies a very high noise level in the data whichmay compromise all steps in time processing.

As a consequence, high demands are posed on theprocessing of sparse seismic data, in order to compensatefor most of the data's short-comings. A suitable processingstrategy should provide a maximum noise suppression

during initial pre-processing and parameter estimation, andin final imaging.

Many conventional strategies increase the fold and signal-quality at the expense of horizontal resolution. The entireprocessing may be based on coarse CMP binning grids thatare initially defined. Similarly, the stacking velocity analy-sis most often requires a temporal combination of severalCMP gathers in supergathers, in order to obtain well-defined and meaningful stacking velocities. This combina-tion of neighbouring CMP gathers, however, fails toimprove the stacking velocity analysis in case of strong dip.The same limitations are found for flexible binning tech-niques that may be used in order to close some data gaps(Spitzer et al., 1998).

Hence, in many situations, coarse processing grids are notadequate for the pre-processing and imaging of sparseseismic data due to the loss of dip information andhorizontal resolution. In the imaging of sparse data,however, a significant dip enhancement and noise suppress-ion have been achieved by the alternative strategy ofCommon-Reflection-Surface, or CRS imaging. Thisimaging technique was successfully applied to sparse datawith very irregular offset distributions and low-fold on finebinning grids (Trappe et al. 2005, Frehers et al. 2007,Eisenberg-Klein et al. 2008).

The CRS method, however, can also be beneficial in thepre-processing of sparse 2D/3D seismic data. In this paper,CRS implications are shown for several pre-processingsteps comprising the choice of the binning grid, the analysisof stacking velocities, and the interpolation of data gaps.

CRS method

The Common-Reflection-Surface, or CRS method wasintroduced by Hubral et al. (1998) as a model-independentimaging technique for zero-offset stacking in time domain.In contrast to NMO zero-offset stacking, the CRS methodassumes a more complex subsurface structure characterizedby reflector elements with dip and curvature. As aconsequence, the corresponding CRS stacking operator isnot limited to a single CMP gather. It collects the reflectionenergy of a subsurface element from all contributing traces,thus including nearby CMPs of the imaging location. Thedefinition of events across several CMP locations alsostabilizes the search for the CRS stacking parameters, orso-called CRS stacking attributes.

Enhanced CRS preprocessing of sparse seismic data

Zero-offset imaging and subsequent migration issignificantly improved by the resulting high-fold CRSstack. Signal-to-noise ratio and reflector continuity areincreased especially in low fold zones (e.g. Trappe et al.2001, Gierse et al. 2003). Partial CRS stacking andCMP/offset regularization of the seismic input datasimilarly provide enhanced prestack data with a strong

noise suppression (Trappe et al. 2008). The resulting so-called CRS gathers are well suited for a stabilized CRS-AVO analysis and prestack imaging in time and depth.CRS stacking and CRS gathers, however, can also have astrong influence on the preprocessing, as is shown in thefollowing examples.

-----------

CRS pseudo 3D processing ofcrooked-line dataA data case of crooked 2D lineacquisition is taken from amountainous terrain that was hardlyaccessible by vehicle beyond sometracks in the main valleys. Inaddition, large areas were stillmarked as explosive ordnance zonesthat were not yet cleared. As aconsequence, acquisition wascarried out in a predominantly 2Dcrooked-line fashion following thetracks (Figure 1).

After an initial 2D crooked line pro-cessing using a bin interval of 25m,possibilities for a 3D processingwere checked. Fine 3D bin cells of25m x 25m showed a low foldaround 2 in most parts of the area,that turned out to be insufficient in aconventional processing. Coarsebins of 50m x 50m increased thefold at the expense of horizontalresolution. In order to preserve aresolution similar to the 2Dprocessing, the fine 3D binning wascombined with the signal enhance-ment of the 3D CRS processing.

Figure 2 shows the stack results onthe fine binning grid fromconventional CMP processing andfrom CRS processing, respectively.The CRS processing shows a strongincrease of the signal-to-noise ratio,and provides a dip consistentextrapolation of the seismic data.Meaningful stacking velocitiescould only be derived fromenhanced CRS gathers provided byCRS partial stacking andregularization.

Figure 1: Fold maps of pseudo 3D processing of crooked line data. Red line indicates location of crosslines from Figure 2

Figure 2: Cross line of CMP stack (left) versus CRS stack (right) of pseudo 3D data

Enhanced CRS preprocessing of sparse seismic data

Velocity analysis on CRS gathersThe derivation of stacking velocitiesis difficult in sparse seismic datawhere CMP gathers comprise only afew traces over large offset ranges.Supergathers from common-offsetbinning and trace stacking in neigh-bouring CMP gathers only partlyimprove the velocity analysis. Mostoften, the combined traces are notsufficient to discriminate primaryreflections from noise.Figure 3 (top) gives an example of aconventional stacking velocityanalysis on CMP supergathers inthis type of low-fold data. For theCMP location considered, it shows agather-based semblance analysis,the associated moveout-correctedsupergather, and five stack panelswith incremental variation of thecentral velocity function. The stackpanels clearly show dipping events.The dip limits the number of CMPlocations that can be combined inthe CMP supergather. At four near-offset traces of the CMP superga-ther no data is available. Moreover,data quality is very low at small andmedium offsets, thus making thevelocity analysis ambiguous.Figure 3 (bottom) shows thealternative use of CRS gathers invelocity analysis. Since CRSprocessing takes the event dips intoaccount, the contributions to theCRS gather can be collected from alarger range of neighbouring CMPgathers. As a result, the seismic datacould be extrapolated to most of thenear-offset traces here. In addition,the partial CRS stacking and dataregularization in incremental offsetranges improve the signal-to-noiserange at all offsets. It should benoted that this partial CRS stackingpreserves residual moveout, e.g. attimes 0.8-0.9s, and irregular move-out, e.g at large offsets. The CRSgathers are thus well-suited forfurther improvements of moveoutflattening and stacking velocityanalysis in sparse and noisy seismicdata.

Figure 3: Stacking velocity analysis on CMP supergathers (top), and CRS gathers (bottom)

Data interpolation by the CRS technique

Sparse 3D data acquisition most often implies strong variations of the offsetdistribution, and corresponding variations of the surface mute. This is obvious inFigure 4 (top), showing a time slice at 700 ms, and a cross section of a 3D CMPstack from sparse data. The time slice shows various data gaps due to surfacemutes in inhabited areas and other types of infrastructure zones where seismicsources were not permitted. The cross section exhibits the corresponding mutezones which partly extend down to 1.2 seconds. Due to these mutes, the subsurfacestructure is rather ill-defined in the upper second of the section.

A 3D CRS processing of this data was carried out with the objective to interpolatethe seismic data into the mute gaps, and to improve the structure at the top of

Enhanced CRS preprocessing of sparse seismic data

Figure 4: Time slices at 700 ms (left) from CMP stack (top), and CRS stack (bottom), with associated crossline sections (right).The red lines indicate the locations of crosslines, and time slice, respectively

the section. Figure 4 (bottom) shows the time slice and thecross section of a 3D CRS stack corresponding to the CMPresults above. The data gaps of the CMP time slice arealmost completely filled in the CRS time slice. In addition,the seismic phases are resolved with better continuity andsignal-to-noise ratio. The CRS crossline shows an almostcomplete data interpolation in the mute gaps up to 0.4s.Moreover, it offers a much clearer view of the anticlinalstructure, both in the interpolated areas and in thepreviously available areas.

Conclusions

The CRS method is a versatile tool for processing 2D and3D sparse seismic data that does not only produce images

of increased signal-to-noise ratio and resolution, but alsoprovides solutions to various preprocessing issues. The dataexamples show that CRS processing allows a fine binningof low-fold data without compromising the image quality,an interpolation of data gaps, and a derivation of reliablestacking parameters.

Acknowledgments

We acknowledge the permission of Chevron, Perenco,TPAO (Turkish Petroleum), and further anonymous oilcompanies, to present their data.