A simulated Simultaneous Source Experiment in Shallow waters and the Impact of Randomization
`Schemes
`Rolf Baardman, Roald van Borselen, PGS
`
`Summary
`
`In simultaneous source acquisition, seismic data can be
`recorded with a temporal overlap between the shots. Better
`sampled data in terms of source spacing, azimuth and/or
`offset distributions can be obtained in a much more
`efficient way. These potential benefits can only be realized
`if the recorded data, with interfering energy from multiple
`sources, can be handled properly. Common practice is to
`apply randomized time-delays to the sources during the
`acquisition of the data. As a result of using randomized
`firing schemes, coherency measures can be utilized to
`actively separate the recorded data over the individual
`sources. In this paper an inversion-based source separation
`method is utilized to a shallow water data set which may
`have specific challenges compared
`to deeper water
`applications. We will focus a bit more on the randomized
`firing schemes. It is shown that optimizing these firing
`schemes, introducing “pseudo randomization”, instead of
`using random time-delays, can benefit the performance of
`the source separation.
`
`The separation method is illustrated using a controlled
`simultaneous source experiment where a shallow water
`field data set is used to mimic simultaneous recorded data
`where two sources were located with only a small cross line
`distance
`between
`them
`(simultaneous FLIP/FLOP
`acquisition). Results demonstrate that it is advised to utilize
`“pseudo randomization” of the firing delay-times. The
`controlled shallow water field data example shows that
`good separation results are obtained.
`
`Introduction
`
`In seismic exploration, there is continuous drive towards
`more dense data sampling to better image complex
`geological structures. Recent advances in acquisition such
`as Wide-Azimuth, Multi-Azimuth or Rich-Azimuth
`acquisition can deliver a more diverse range of source,
`azimuth and offset sampling. To collect such data, multiple
`source and
`receiver vessels are deployed,
`thereby
`increasing the costs of the survey significantly.
`
`In conventional acquisition, there is zero time overlap
`between
`shot
`records,
`and
`data
`are
`recorded
`discontinuously. The source domain
`is often poorly
`sampled, leading to aliasing.
`
`recorded
`In simultaneous acquisition, data can be
`continuously, and temporal overlap between shots is
`allowed. Consequently, more sources are fired during the
`
`
`
`
`
`where D is the blended data matrix, zd and zs are the
`detector and source depth level respectively. Blending
`matrix Г
`(Berkhout 2008) contains
`the blending
`parameters. In the case of a marine survey with random
`firing times but equal source strengths, only phase
`encoding is utilized. As such, elements Гkl from the
`blending elements only consist of phase terms exp(─ jωτkl)
`
`© 2013 SEG
`SEG Houston 2013 Annual Meeting
`
`DOI http://dx.doi.org/10.1190/segam2013-1199.1
`Page 4382
`
`same period of acquisition, which greatly enhances the
`flexibility in survey geometries. As a result, a more densely
`sampled data set in terms of source spacing, but also
`azimuth and offset distributions can be obtained. In terms
`of efficiency, simultaneous acquisition can contribute by
`reducing survey times, which is of particular value in
`critical situations where small acquisition time-windows
`dominate due to severe safety, environmental or economic
`restrictions.
`
`As such, from an acquisition point of view, simultaneous
`acquisition holds the promise of both efficiency and quality
`improvements. However, unless source separation can be
`achieved to a sufficiently high degree, the enormous
`potential benefits of
`simultaneous
`sources
`remain
`unrealized.
`
`In this abstract, an inversion-driven method is utilized that
`aims to distribute all energy in the blended shot records by
`reconstructing the individual unblended shot records at
`their respective locations. The focus is this paper will be on
`shallow water applications. The method is explained further
`in the next section, after which we discuss how the firing
`schemes can be optimized and finally a controlled field
`data examples is presented.
`
`Methodology
`
`Inversion-driven methods aim to construct the separated
`sources through the minimization of a cost function that
`describes the “data misfit” (see, for example, Akerberg et
`al. 2008 and Moore et al. 2008).
`
`Using the well-known matrix notation (Berkhout 1982),
`seismic data in the temporal frequency domain can be
`represented by data matrix P, where each element
`corresponds to a complex-valued frequency component of a
`recorded trace, the columns representing shot records and
`the rows receiver gathers. In general, source blending can
`be formulated as follows:
`
`D (zd, zs ) = P (zd, zs ) Г
`
`
`
` (1)
`
`Downloaded 09/18/15 to 64.124.209.76. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
`
`PGS Exhibit 2017
`WesternGeco v. PGS (IPR2015-00309, 310, 311)
`
`

`
`A simulated Simultaneous Source Experiment in Shallow waters and the Impact of Randomization Schemes
`
`
`
`
`
`
`A
`
`B
`
`
`
`
`
`
`
`Figure 1: Common channel gather for recorded simultaneous
`source data (A) and its separation result (B). Notice that (indicated
`in the red box), the separation result shows leakage because
`interfering energy, that should be incoherent in this domain, is
`misinterpreted as coherent energy for the wrong source.
`
`
`
`Acquisition and operator window size of the coherency
`filters can be used to determine a minimal number of
`adjacent shots (both inline and cross line) for which
`randomness should be secured. Consider a simple
`acquisition with 1 recording streamer, 2 simultaneous
`sources and a coherency filter with an operator length of 20
`traces. Considering that only one of the sources is
`randomized (other will always fire at t=0), one should
`make sure that within 20 adjacent shots no time-delays are
`the same or close to each other for the randomized source.
`One way to do so is to divide the total time range of
`allowed time-delays (for instance 0 -1000ms) into 20
`groups (group1: 0-50ms, group2: 50-100ms … group20:
`950-1000ms). Pseudo randomize the order of the groups
`(group7, group15, group3 ….). For the first shot number, a
`time-delay
`is picked
`from
`the
`first group after
`randomization (group7 in this case) and applied to the
`randomized source. For the second shot number, a time-
`delay for the randomized source is picked from the second
`group after randomization (group 15 in this case). When
`shot number 21 is reached, the first group (group7) is used
`again to pick a time-delay. This way it is possible to ensure
`that there is enough randomness within the operator
`window to avoid leakage as shown in Figure1. In case more
`sources are utilized in a simultaneous source experiment
`(and multiple sources are randomized), it is proposed to
`first determine the “random seed”, the delay-times of all
`simultaneous sources for shot number 1. Then, define one
`of the simultaneous sources as reference source and use the
`system described above to determine the delay-time for this
`reference source for the second shot number. Change the
`delay-times for all other sources with the same amount that
`the delay time for the reference source was changed.
`
`
`
`that express the time delay τkl given to source k in blended
`source array l.
`
`To retrieve individual „deblended‟ shot records from
`blended data, a matrix inversion has to be performed. In
`general, the blending problem is underdetermined meaning
`that there is no unique solution to the inverse problem.
`Hence, the blending matrix is not invertible.
`
`
`In this paper, an inversion based separation method
`(Baardman and van Borselen 2012, van Borselen et al.
`2012) is used that constrains the inversion based on
`coherency measures (Abma et al. 2010). The method is
`utilized in a mixed common channel / CDP domain. The
`randomized time-delays applied to the sources during the
`acquisition ensures that, dependant for which source you
`align the data, energy for one source will become coherent
`while all interfering energy from other sources appear as
`incoherent spikes. In an iterative way all the individual
`separated gathers are build up simultaneously. In each
`iteration a multi-dimensional median filter extracts the
`strongest component of coherent energy for all individual
`sources. Advantage is that when the strong events are
`separated first, the weaker events are better accessible and
`can be better separated.
`
`
`
`Optimized design of firing scheme
`
`
`Since the separation method is based on coherency
`measures it is of vital importance that the randomized time-
`delays applied to the sources ensures that there is enough
`randomness in the utilized domain(s). In case random
`numbers are generated, it may occur that, within a couple
`of shots, two or more shots have time-delays that are the
`same or very similar. Energy from that source, that should
`appear as incoherent spikes in a gather aligned for another
`source, can now be misinterpreted as coherent energy for
`the wrong source resulting in leakage. Figure 1a shows a
`common offset gather of a simultaneous field data example
`where, because random time-delays were generated, 3 out
`of 4 adjacent shots had accidently almost identical time-
`delays applied to them. As a result, the interfering energy,
`that should be incoherent, can now easily be misinterpreted
`and leak into the wrong source. Figure 1b shows the
`separation result for the same gather and indeed we see that
`energy has leaked to the wrong source.
`
`Instead of generating the time-delays randomly, it is
`proposed to do that pseudo-randomly where a priori
`information of the acquisition, operator window size and
`geology can be used to constrain the process.
`
`
`
`© 2013 SEG
`SEG Houston 2013 Annual Meeting
`
`DOI http://dx.doi.org/10.1190/segam2013-1199.1
`Page 4383
`
`Downloaded 09/18/15 to 64.124.209.76. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
`
`PGS Exhibit 2017
`WesternGeco v. PGS (IPR2015-00309, 310, 311)
`
`

`
`A simulated Simultaneous Source Experiment in Shallow waters and the Impact of Randomization Schemes
`
`
`A
`
`B
`
`C
`
`
`
`Shallow water data example
`
`In this example, the proposed method is deployed using a
`controlled simultaneous source experiment using a shallow
`water field data set from offshore UK.
`In shallow water the following challenges may occur:
`- Presence of high amplitude refracted energy
`- Presence of many short-period surface multiples
`
`The field data set is blended manually; time shift between
`250 – 1000ms are applied to the shots and added to the
`original data set.
`
`A
`
`B
`
`C
`
`D
`
`
`
`Figure 2: Shot gather of A) Blended input data, B) Separation
`result for source 1 C) Residual energy and D) Difference between
`separation result and reference data for source 1.
`
`Figure 3: Common near offset channel of A) Blended input data,
`B) Separation result for source 1 C) Difference between separation
`result and reference data for source 1.
`
`© 2013 SEG
`SEG Houston 2013 Annual Meeting
`
`DOI http://dx.doi.org/10.1190/segam2013-1199.1
`Page 4384
`
`Downloaded 09/18/15 to 64.124.209.76. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
`
`PGS Exhibit 2017
`WesternGeco v. PGS (IPR2015-00309, 310, 311)
`
`

`
`A simulated Simultaneous Source Experiment in Shallow waters and the Impact of Randomization Schemes
`
`This way, we simulate a simultaneous source experiment
`with 2 sources, one always fired at zero time and one
`randomized. Because it is a controlled experiment we can
`compare the separation results to the optimal separation
`result, the reference data. Figure 2a shows an arbitrary shot
`record of the blended input data. In Figure 2b the
`separation result for source 1 is plotted. Figures 2c,d show
`the residual energy after separation and the difference
`between the separation results and the reference data for
`source 1. Similar separation results were obtained for the
`second source. Figure 3a shows a common near offset
`channel for the blended input data. The separation result for
`source 1 is shown in Figure 3b. Note the very good signal
`preservation of the events after separation, retaining all
`events optimally. The difference plot to the reference data
`is plotted in Figures 3c. The absence of coherent energy
`shows again the good signal preservation while the residual
`interfering noise from the secondary source is limited. Note
`also that no additional filtering was applied to achieve these
`results: only the inversion-based source separation method
`was utilized. Figure 4 show some stacked sections of the
`separation results for source 1. In Figure 4a the blended
`input data, aligned for source 1, is plotted. Figure 4b,c
`show the separation result and difference to the reference
`data for source 1. Similar conclusions can be drawn from
`these results; good signal preservation with acceptable
`residual noise level is achieved.
`
`
`
`Conclusions
`
`In this paper we revisited an inversion-based source
`separation approach. The use of randomized firing schemes
`in the acquisition allows the method to utilize coherency
`criteria to solve the source separation inverse problem. It is
`shown that generating the time-delays pseudo-randomly
`instead of randomly, will benefit the separation process.
`With random time-delays the possibility is not excluded
`that interfering energy, that should appear as incoherent
`spikes, can accidently be misinterpreted as coherent energy
`for the wrong source. Selecting the time-delays pseudo-
`randomly using minimal a priori information helps to
`prevent leakage of this kind. Results from a controlled
`shallow water field data experiment indicate that the
`separation performs very well. The challenges with shallow
`water do not seem to be an issue in this particular
`application. Very good signal preservation is achieved with
`minimal residual energy from the interfering sources.
`
`
`
`
`
`A
`
`B
`
`C
`
`Figure 4: Stacked section of A) Blended input data, B) Separation
`result for source 1 C) Difference between separation result and
`reference data for source 1.
`
`
`© 2013 SEG
`SEG Houston 2013 Annual Meeting
`
`DOI http://dx.doi.org/10.1190/segam2013-1199.1
`Page 4385
`
`Downloaded 09/18/15 to 64.124.209.76. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
`
`PGS Exhibit 2017
`WesternGeco v. PGS (IPR2015-00309, 310, 311)
`
`

`
`http://dx.doi.org/10.1190/segam2013-1199.1
`
`EDITED REFERENCES
`Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2013
`SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for
`each paper will achieve a high degree of linking to cited sources that appear on the Web.
`
`REFERENCES
`Abma, R., T. Manning, M. Tanis, J. Yu, and M. Foster, 2010, High-quality separation of simultaneous
`sources by sparse inversion: 72nd Annual International Conference and Exhibition, EAGE, Extended
`Abstracts, B003.
`Akerberg, P., G. Hampson, J. Rickett, H. Martin, and J. Cole, 2008, Simultaneous source separation by
`sparse Radon transform: 78th Annual International Meeting, SEG, Expanded Abstracts, 2801–2805.
`Baardman, R., and R. van Borselen, 2012, Separating sources in marine simultaneous shooting
`acquisition – Method and applications: 82nd Annual International Meeting, SEG, Expanded
`Abstracts, 1–5
`Berkhout, A. J., 1982, Seismic migration: Imaging of acoustic energy by wave field extrapolation — A:
`Theoretical aspects: Elsevier Scientific Publishing Company.
`Berkhout, A. J., 2008, Changing the mindset in seismic data acquisition: The Leading Edge, 27, 924–938,
`http://dx.doi.org/10.1190/1.2954035.
`Moore, I., B. Dragoset, T. Ommundsen, D. Wilson, C. Ward, and D. Eke, 2008, Simultaneous source
`separation using dithered sources: 78th Annual International Meeting, SEG, Expanded Abstracts,
`2806–2810.
`Van Borselen, R., R. Baardman, T. Martin, B. Goswani, and E. Fromyr, 2012, An inversion approach to
`separating sources in marine simultaneous shooting acquisition — Application to Gulf of Mexico
`data set: Geophysical Prospecting, 60, no. 4, 640–647, http://dx.doi.org/10.1111/j.1365-
`2478.2012.01076.x.
`
`
`
`© 2013 SEG
`SEG Houston 2013 Annual Meeting
`
`DOI http://dx.doi.org/10.1190/segam2013-1199.1
`Page 4386
`
`Downloaded 09/18/15 to 64.124.209.76. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
`
`PGS Exhibit 2017
`WesternGeco v. PGS (IPR2015-00309, 310, 311)