Th ELI 9 Brodnd Processing of West of Shetlnd Dt R. Telling* (Dolphin Geophysicl Limited), N. Riddlls (Dolphin Geophysicl Ltd), A. Azmi (Dolphin Geophysicl Ltd), S. Grion (Dolphin Geophysicl Ltd) & G. Willims (Dolphin Geophysicl Ltd) SUMMARY We present rodnd processing results for dtset cquired north-west of Shetlnd. Attenution of multiples ws chllenging ecuse of the presence of hrd, vrile depth wter ottom nd complex geology. Stremers were towed deep (m to m) during cquisition, with mild slnt of m per km of offset, nd with mximum offset of km. This cquisition configurtion llowed the use of multiple ttenution methods designed for conventionl dt, while t the sme time llowing rodnd postmigrtion de-ghosting with n overll processing turnround time comprle to tht for conventionl processing projects. 7 th EAGE Conference & Exhiition Amsterdm RAI, The Netherlnds, -9 June
Introduction This pper descries the rodnd processing of deep-tow D mrine survey dt otined s prt of rodnd regionl Froes/West of Shetlnd well-tie Survey. A centrl feture of the processing ws the ility to otin finl rodnd product using wht is essentilly conventionl pre-migrtion processing sequence. The only ddition to the sequence is step for receiver-side ghost removl fter imging. The use of stndrd de-multiple nd pre-stck time migrtion lgorithms contrsts with the configurtion-specific processing lgorithms (Slon et l. ) or pre-imging de-ghosting tht re necessry when rodnd cquisition utilizes strongly curved or slnted cle. A series of exmple imges re provided to illustrte key steps in the processing sequence. Dt cquisition Fifteen D lines totlling km were cquired in the Shetlnd-Froe sin in My y MV Artemis Atlntic. The source ws n irgun rry towed t 8m depth, with totl volume.7m ( cuic inches) nd pressure r (psi). The shot intervl ws m. The receiver cle ws towed deep with very gentle slnt of m per km to fcilitte rigging nd minimize tug noise. Cle depth ws m t the ner end nd m t the fr end. The cle comprised 8 hydrophone chnnels, the first t offset m nd with chnnel spcing of.m, so km length in totl. The filters used were Hz-8dB/octve nd Hz-7dB/octve. The exmple line shown in this pper ws 9km in length, cquired cross the sin siling in north-westerly direction. Processing sequence The processing sequence in outline is outlined in Tle. The most prominent multiples oserved in the dtset were wter-lyer multiples originting from the high impednce contrst primry reflectors. These were the wter ottom nd of the top of the slt lyer. These key primries re mrked in Figure s WB nd TB respectively together with exmples of their respective order of multiples, mrked with suffix M nd M.. Re-dtum to men se-level 8. Offset regulriztion. Source de-signture 9. F-X deconvolution. Anti-lis filter nd resmple to ms. D Kirchoff pre-stck time migrtion. Swell nd liner noise ttenution. Hi-resolution de-multiple. Multiple ttenution. De-ghosting. Anti-lis nd lternte chnnel drop. Post-stck deconvolution nd scling 7. Hi-resolution Rdon de-multiple Tle Processing sequence. After dtum correction to men se-level, ttenution of multiples ws crried out using stndrd implementtion of D SRME. Figure shows the result for our exmple line nd it cn e seen tht SRME worked very well in the mjority of res; the wter column eing in the rnge m to m. However, in the shllower regions of the survey, e.g. the yellow ox in Figure, it cn e seen tht there is some residul energy tht needs ttention. Here, the ccurcy of the SRME model is confounded y n incomplete record of the seed reflection t the ner offsets, nd y multiples tht re very high in mplitude compred to the primries (due to the hrd seed). We therefore investigted dditionl multiple models (Figure ). The first ws produced using wve-field extrpoltion lgorithm, which complemented the SRME result for the multiples ssocited with the deeper horizons nd the second ws generted using n lgorithm designed specificlly for shllowwter multiple ttenution. This ltter pproch derives n opertor from the dt tht represents the (missing) wter-ottom reflection nd susequently convolves this with the originl dt to generte multiple model. See for exmple Hrgreves () for discussion on this type of lgorithm. 7 th EAGE Conference & Exhiition Amsterdm RAI, The Netherlnds, -9 June
WB M WB WB M TB WB M TB M WB M WB M WB WB M See Figure TB WB M TB M WB M Figure Stcked section efore nd fter ppliction of D SRME which shows generlly very effective ttenution of multiple energy. WB = wter-ottom, TB = top of slt, M#=multiple, nth order. However, note there is residul surfce multiple energy corresponding to the TB horizon t the right-hnd-side of the section, outlined with the yellow ox. In ll cses, the multiple model is dptively sutrcted from the originl dt with creful ttention pid to preserving primry signl while mximizing multiple removl. In the shllow region, much improved result over SRME lone ws found fter second stge of sutrction using the multiple model derived from wve-field extrpoltion, Figure c. However, it ws found possile to otin comprle ttenution using the shllow wter lgorithm lone, Figure d. This pproch ws lso found to e more effective t removing the first nd second order multiples from the wter-ottom nd eneth the centrl elt of slt. The simpler methodology nd improved result fvour this choice of lgorithm for the dtset. After velocity picking, further ttenution of multiple energy ws crried out using hi-resolution prolic Rdon trnsforms with mute to discriminte primry from multiple sed on residul move-out. After de-multiple, lternte chnnels were dropped, the offsets were regulrized, treted gin for noise, nd the inned offsets migrted using -D Kirchhoff prestck time migrtion lgorithm with km perture nd degree dip limit. The migrted stck with zero-phse wvelet is provided in Figure. After finl round of residul noise nd multiple removl, the migrted pre-stck dtset ws deghosted. The high signl to noise rtio tht is key to successful ghost removl ws chieved y towing stremers deep in order to reduce the effects of swell noise. The receiver nd source ghost notches induced y se-surfce reflections were compensted for y post-imging D de-ghosting process, pplied using D opertor. The de-ghosting process ws time-spce vrint, nd de-ghosting opertors were derived using comintion of dptive estimtion nd -priori informtion on cle depth, wter velocity nd se-surfce reflection coefficient. The process lso included sttisticl 7th EAGE Conference & Exhiition Amsterdm RAI, The Netherlnds, -9 June
corrections for the minimiztion of ny residul ghost energy. The finl de-ghosted time-migrted stck is shown in Figure. c d Figure Stcked section for - seconds, showing zoomed nd dditionlly gined section fter figure. is input efore multiple ttenution is fter D SRME c is fter SRME nd use of the wve-field extrpoltion de-multiple model, finlly c is the shllow wter de-multiple lgorithm pplied to input for comprison. Discussion The success of rodnd processing requires the removl of ghost events from seismic dt. These cuse frequency-dependent constructive nd destructive interference, with consequent ttenution of signl in specific frequency nds (notches). Vrile depth cquisitions with either curved or pronounced liner slnt provide notch diversity for stcking purposes. However, they induce strong vriility in ghost response with respect to offset, therefore reking down the sttionry wvelet ssumption tht mny processing lgorithms require (e.g. Semlnce velocity nlysis, SRME). A more complex, time consuming processing sequence is therefore required (Slon et l., ), or pre-imging de-ghosting my e needed in order to keep the processing sequence simple. On the other hnd, when the cle depth is kept constnt the processing sequence prior to de-ghosting does not require modifiction, nd processing turnround time is not ffected. Additionlly, in the cse of constnt cle depth cquisitions, de-ghosting does not rely on notch diversity ut on signl to noise rtio (SNR), s discussed in Willims nd Polltos (). The cle is towed deep for ll offsets, therefore gurnteeing high SNR even t the ner offsets, in contrst with slnted or curved cquisitions where these offsets re shllower. Grion et l. () compred rodnd imges over djcent lines cquired in clm nd rough ses, thus confirming tht deep tows provide isoltion from wether effects for structurl imging purposes. De-ghosting is then performed efore stcking migrted offsets. For opertionl resons, the cle my e towed with slight slnt (e.g. m per km of offset), which still llows for conventionl pre-de-ghosting processing sequence. 7th EAGE Conference & Exhiition Amsterdm RAI, The Netherlnds, -9 June
Figure Time-migrted imges efore nd fter de-ghosting. Conclusions Inspection of migrted imges nd NMO stcks hs demonstrted tht successful multiple removl cn e chieved for chllenging rodnd West of Shetlnd dtset. Algorithms designed for conventionl seismic dt re therefore confirmed to e suitle for rodnd processing, when cquisition is performed using constnt depth deep tow, or with mild slnt. For the dtset here presented, slnt of m per km of offset ws used, for opertionl resons only. Acknowledgements The dt shown re provided courtesy of Dolphin's Multi-Client Deprtment. References Grion, S., Azmi, A., Polltos, J., Riddlls, N. nd Willims, R.G. [] Brodnd processing with clm nd rough ses: oservtions from North Se survey. 8rd Annul SEG Meeting, Expnded Astrcts, -. Hrgreves, N. [] Surfce multiple ttenution in shllow wter nd the construction of primries from multiples. 7th Annul SEG Meeting, Expnded Astrcts, 89-9. Slon, R., Russier, R., Hrdouin, D., Grtcos, B., Sours, R. nd Lin, D. [] Multiples ttenution for vrile depth stremer dt, shllow nd deep wter cses. EAGE Extended Astrcts, Y. Willims, R.G. nd Polltos, J. [] The key to incresed mrine seismic ndwidth. First Brek,, -. 7th EAGE Conference & Exhiition Amsterdm RAI, The Netherlnds, -9 June