Method and system for generating geophysical data

Abstract

A method of generating geophysical data using at least one source. The method may include the steps of generating a geophysical wavefield with a varying signature using at least one source, wherein the signature is varied in a periodic pattern.

Claims

1. A method of generating geophysical data using at least one source, the method comprising: generating a geophysical wavefield with a varying signature using at least one source, wherein the signature is varied in a periodic pattern; recording geophysical energy to produce geophysical data using at least one receiver, the geophysical energy comprising the propagating geophysical wavefield generated at the at least one source; and transforming the geophysical data into another domain; wherein the other domain is a domain such that at least some of the geophysical data is shifted to a location that is different to the location in the other domain where the at least some of the geophysical data would have been had the varying signature not been used.

2. The method as claimed in claim 1, wherein the at least one source is an airgun source, an airgun source array, a watergun source, a flip/flop source, or an electric and/or magnetic source.

3. The method as claimed in claim 1, further comprising: isolating the geophysical data originating from the propagating geophysical wavefield generated at the at least one source from any other geophysical data that may be present in the other domain.

4. The method as claimed in claim 3, further comprising transforming the isolated geophysical data back into the domain in which the geophysical data was recorded.

5. The method as claimed in claim 3, further comprising: conditioning the isolated data so as to effectively remove the varying signature pattern from the recorded geophysical data.

6. The method as claimed in claim 5, wherein the conditioning step occurs in the other domain or in the domain in which the geophysical data was recorded.

7. The method as claimed in claim 1, further comprising: recording geophysical energy to produce geophysical data using at least one receiver, the geophysical energy comprising the propagating geophysical wavefield generated at the at least one source; and isolating the geophysical data originating from the propagating geophysical wavefield generated at the at least one source from any other geophysical data that may be present in the other domain.

8. The method as claimed in claim 1, wherein the periodic pattern is such that, after transforming the recorded geophysical data into another appropriate domain, a first portion of the recorded geophysical data originating from the propagating geophysical wavefield generated by the at least one source would be shifted relative to a second portion of the recorded geophysical data originating from the propagating geophysical wavefield generated by the at least one source, and the method comprises: identifying the first portion; and processing the data to calculate a full data signal at the shifted location of the first portion using the identified first portion and/or to remove the second portion of the data using the identified first portion.

9. The method as claimed in claim 1, wherein the signature is varied using time dither and the varying time dither is that every second geophysical wavefield generated by the at least one source is triggered with a constant delay of time T.

10. The method as claimed in claim 1, wherein the signature is varied by varying the polarity and the varying polarity is that every second geophysical wavefield generated by the at least one source has opposite polarity.

11. The method as claimed in claim 1, comprising selecting the varying signature of the at least one source such that, once geophysical energy comprising the generated geophysical wavefield and another signal is recorded and the recorded geophysical data is transformed into another appropriate domain, the recorded geophysical data originating from the generated geophysical wavefield would be shifted away from recorded geophysical data originating from the other signal.

12. The method as claimed in claim 11, wherein the other signal arises from noise, interference, or one or more other sources.

13. The method as claimed in claim 1, wherein at least two sources are used to simultaneously generate geophysical wavefields, the first source having a varying signature in a periodic and the second source having no varying signature in a periodic, or having a different varying signature in a periodic pattern and/or wherein the method comprises selecting the varying signature such that, once the geophysical data is recorded and transformed into another domain, a pressure wave portion of the geophysical data will be at least partially shifted away from a shear wave portion of the geophysical data.

14. The method as claimed in claim 1, the method comprising selecting the varying signature such that, once the recorded geophysical data is recorded and transformed into another domain, the portion of the recorded geophysical data originating from the generated wavefield would be at least partially shifted away from an interference portion of the recorded geophysical data.

15. The method as claimed in claim 14, wherein the signature is varied using time dither, and wherein the interference portion has a dominant frequency, and the method comprises using a time dither of approximately the same as, a half of or a quarter of the period of the dominant frequency.

16. The method as claimed in claim 1, the method comprising selecting the varying signature such that, once the geophysical data is recorded and transformed into another domain, a residual shot noise portion of the recorded geophysical data would be at least partially shifted away from the portion of the geophysical data originating from the generated geophysical wavefield.

17. The method as claimed in claim 16, wherein the signature is varied using time dither, and wherein the residual shot noise portion has a dominant frequency, and the method comprises using a time dither of approximately the same as, a half of or a quarter of the period of the dominant frequency of the residual shot noise.

18. The method as claimed in claim 16, wherein the signature is varied by varying the polarity, and wherein the residual shot noise portion has a dominant frequency, and the periodic pattern of the varying polarity of sequentially generated geophysical wavefields is: a second generated geophysical wavefield having the same polarity as a first generated geophysical wavefield, a third generated geophysical wavefield having opposite polarity to the second generated geophysical wavefield, a fourth generated geophysical wavefield having the same polarity as the third generated geophysical wavefield, a fifth generated geophysical wavefield having opposite polarity to the fourth generated geophysical wavefield, a sixth generated geophysical wavefield having the same polarity as the fifth generated geophysical wavefield, (i.e. +1, +1, −1, −1, +1, +1, −1, −1).

19. The method as claimed in claim 1, wherein the time between generating subsequent geophysical wavefields is less than the time taken for the geophysical wavefield energy originating from each generated geophysical wavefield to be recorded by the receiver, the method comprising identifying the data in a given trace originating from a geophysical wavefield generated previously to the trigger time of the given trace, and adding this identified data to data on a previous trace originating from the same geophysical wavefield.

20. The method as claimed in claim 1, comprising reducing the width of the data signal originating from the at least one source in the other domain.

21. The method as claimed in claim 20, when the geophysical wavefield, energy and/or data is a seismic wavefield, energy and/or data, comprising removing low-speed waves of the recorded wavefield.

22. The method as claimed in claim 1 used in a modeling, imaging or inversion method; and/or wherein the geophysical data is 2D or 3D geophysical data; and/or wherein the geophysical wavefield, energy and/or data is a seismic wavefield, energy and/or data, or the geophysical wavefield, energy and/or data is a controlled source electromagnetic wavefield, energy and/or data; and/or wherein the transform may be a Fourier, tau-p or radon transform.

23. A system for generating geophysical data comprising: at least one source for generating a geophysical wavefield with a varying signature, wherein the source is configured to vary the signature of the geophysical wavefield in a periodic pattern; at least one receiver for recording geophysical energy, the geophysical energy comprising the propagating geophysical wavefield generated at the at least one source; and a processor for transforming the recorded geophysical data into another domain, wherein the other domain is a domain such that at least some of the recorded geophysical data is shifted to a location that is different to the location in the other domain where the at least some of the geophysical data would have been had the varying signature not been used.

24. The system as claimed in claim 23, comprising: a processor for isolating the geophysical data originating from the propagating geophysical wavefield generated at the at least one source from any other geophysical data that may be present in the other domain, wherein the processor comprises a filter for filtering the recorded data.

25. The system as claimed in claim 23, comprising at least two sources each for generating a geophysical wavefield, the first source being configured to vary the signature of its geophysical wavefield in a periodic pattern, and the second source being configured not to vary the signature of its geophysical wavefield in a periodic pattern, or configured to vary the signature of its geophysical wavefield in a different periodic pattern.

26. The system as claimed in claim 23 configured to perform the method of claim 1.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Preferred embodiments of the invention will now be discussed, by way of example only, with reference to the accompanying drawings, in which:

(2) FIG. 1 shows an illustration of what a common receiver gather seismic data set may look like in fk after conventional shooting (left) and flipping polarity on every second shot (right).

(3) FIG. 2 shows an illustration of what a common receiver gather seismic data set may look like in fk after using a time dither on every second shot.

(4) FIG. 3 shows an example acquired data set used to illustrate an embodiment of the present invention.

(5) FIG. 4 shows the fk spectrum of the data shown in FIG. 3.

(6) FIG. 5 shows the fk spectrum of the data shown in FIG. 3, but where the polarity of every second trace has been flipped, emulating a survey where every second shot has opposite polarity to the shots right before and after.

(7) FIG. 6 shows the fk spectrum of the data shown in FIG. 3 but flipping polarity on traces such in the pattern +1, +1, −1, −1, +1, +1, −1, −1, etc., emulating a survey where the polarity of shots are flipped in that pattern.

(8) FIG. 7 shows the fk spectrum of the data shown in FIG. 3 but where a time shift of 10 ms has been applied to every second trace, emulating a survey where a time dither of 10 ms on every second shot is used.

(9) FIG. 8 shows the fk spectrum of the data shown in FIG. 3 but where a time shift of 20 ms has been applied to every second trace, emulating a survey where a time dither of 20 ms on every second shot is used.

(10) FIG. 9 shows the fk spectrum of the data shown in FIG. 3 but where a time shift of 40 ms has been applied to every second trace, emulating a survey where a time dither of 40 ms on every second shot is used.

(11) FIG. 10 shows the fk spectrum of the data shown in FIG. 3 but where a time shift of 200 ms has been applied to every second trace, emulating a survey where a time dither of 200 ms on every second shot is used.

(12) FIG. 11 shows the data of FIG. 7 where equation 17 has been used to properly “deghost” the signal cone centred around the Nyquist wavenumber so that the shifted data now again corresponds to the original data (but shifted to the Nyquist wavenumber).

(13) FIG. 12 shows the data shown in FIG. 11 where equation 17 has been used to properly “reghost” the signal cone centred around the Nyquist wavenumber and subtract it from the signal cone centred around k=0. We are left with a data set with one signal cone only that has been shifted to the Nyquist wavenumber.

(14) FIG. 13 shows a possible configuration for a flip/flop source to achieve a spatially fully uniform distribution of shot points when time dither is used.

(15) FIG. 14 shows the effect on signal and residual shot noise arrivals after backing off time shifts.

DETAILED DESCRIPTION

(16) In one embodiment, the present method relates to a new way to acquire seismic data based on how seismic sources are utilized. The key realization is that by varying the source signature from shot to shot it is possible to separate data from other signals or noise. In one embodiment of the invention a source boat shoots every second shot with a certain source signature while every intermediate shot is fired with the same source signature but with opposite polarity. After a frequency-wavenumber (fk) transform of such data, for instance sorted into a common receiver gather, the data will populate opposite ends of the k axis in the fk spectrum compared to where a conventionally shot data set with the same source signature for every shot would end up. The theory of this is discussed in detail below. We can exploit this effect for a number of different applications, which are each discussed in greater below:

(17) 1. Simultaneous source acquisition. The technique described here offers a new way to acquire simultaneous source data. Two or more sources can be fired simultaneously, and the data received from them can be separated, through the use of different varying signature patterns.

(18) 2. Seismic interference cancellation. By using the present method we can adapt the recorded data signal so as to shift the data originating from the one or more sources from seismic interference. To achieve this, the optimal signature pattern can be adapted as the measurements are taken.

(19) 3. Residual shot noise attenuation. By choosing the signature variation pattern appropriately, we can use the methodology to isolate residual shot noise that can be removed without affecting the signal at all. Benefits include better signal-to-noise ratio in the acquired data, faster acquisition of seismic data, denser acquisition of seismic data, better low frequency acquisition (low frequencies tend to be most affected by residual shot noise).

(20) 4. Seismic data modelling and reverse time migration (RTM). Since the method allows multiple sources to be used simultaneously, there can be dramatic efficiency savings when modelling. For instance, if two sources are used, there are immediate efficiency savings are up to a factor of 2.

(21) 5. Broad-band seismic acquisition. As is discussed further below, once shifted, there may be some aliasing of the data from multiple sources. However, data from multiple sources are always unaliased for low frequencies using our method. We can use purpose built low frequency sources and guarantee that these will not interfere with the acquisition of conventional data.

(22) 6. Cost-effective acquisition of shear-wave data. Similarly to the RSN application, by choosing the signature variation pattern appropriately, we can use the methodology to isolate the shear-wave arrival from the pressure-wave arrival.

(23) 7. Deghosting and source-side gradients for interpolation. By applying the varying source signature pattern to different sub-arrays within an airgun array, it is possible to separate the responses from sub-arrays such that horizontal gradients on the source side can be computed. These are useful for source-side deghosting and other applications.
Theory

(24) In the following discussion of the theory behind the present method, techniques that exploit the fact that the fk space in marine seismic data contains significant portions that are empty limited by apparent propagation velocities that cannot be lower than the propagation velocity in water are discussed. However, other domains and geophysical data types may also be used.

(25) The left part of FIG. 1 illustrates a frequency-wavenumber (fk) plot of, for instance, a common shot gather or common-receiver gather from a marine seismic survey. All signal energy sits inside a “signal cone”. This is because the slowest possible apparent velocity of any seismic energy will correspond to the propagation velocity of water. Outside this signal cone the data is zero in the fk plot.

(26) The inventors have found that by varying the source signature form shot to shot thereby introducing different shooting patterns it is possible to make much better use of the available fk space. The data may be deliberately aliased.

(27) One example of such a shooting pattern is to shoot all even shot points with a certain source signature and interleave all odd shot points using the same source signature but with opposite polarity. For such a data set, a recorded common receiver gather will have every second trace with flipped polarity, or in other words, the following modulating function has been applied to a conventional data set where all traces had the same source signature:
g.sub.1(n)=(−1).sup.n.  (12)

(28) Equation 12 can also be written as
g.sub.1(n)=e.sup.iπn.  (13)

(29) By applying the function g.sub.1 in equation 13 as a modulating function to conventionally recorded (i.e. data recorded without using a varying source signature) data f(n), where n is trace number, before taking a (normalized) discrete Fourier transform:
(f(n))=F(e.sup.ik),

(30) we obtain
(f(n)g.sub.1(n))=(f(n)e.sup.iπn)=F(e.sup.i(k-π)),  (14)

(31) which is a standard Fourier transform result (wavenumber shift). That is, modulating a function with equation 12 results in a wavenumber shift by the Nyquist wavenumber.

(32) The right part of FIG. 1 shows what such a data set would look like after an fk transform. Note that the signal cone has now been shifted laterally so that it is centred at the Nyquist wavenumber k.sub.N with half the signal cone on the negative side of the wavenumber axis and the other half on the positive side.

(33) Next we consider the case that we refer to as time dither. In such a case, every second trace may have a time dither T compared to neighbouring traces. The modulating function that we wish to apply can be written as a superposition of several functions with known transforms:

(34) $\begin{array}{cc}{g}_2\left(n\right)=\frac{1}{2}{\left(-1\right)}^n+\frac{1}{2}-\frac{1}{2}{\left(-1\right)}^n{e}^{i\omega T}+\frac{1}{2}{e}^{i\omega T}.& \left(15\right)\end{array}$

(35) Note that the exponentials are due to Fourier transforms in a different dimension (Fourier transforms of a time shift T) and are constants in the (space) dimension that we consider.

(36) Equation 15 can be written more compact as, the sum of two modulating functions (one of which is a constant with respect to trace number n):

(37) $\begin{array}{cc}{g}_2\left(n\right)=\frac{1}{2}\left[1+e^{i\omega T}\right]+\frac{1}{2}\left[1-e^{i\omega T}\right]{\left(-1\right)}^n.& \left(16\right)\end{array}$

(38) Finally, we can obtain the result:

(39) 0$\begin{array}{cc}ℱ\left(f\left(n\right)g_2\left(n\right)\right)=\frac{1}{2}\left[1+e^{i\omega T}\right]F\left(e^{\mathrm{ik}}\right)+\frac{1}{2}\left[1-e^{i\omega T}\right]F\left(e^{i\left(k-\pi \right)}\right).& \left(17\right)\end{array}$

(40) Equation 17 shows that the seismic data will be mapped in two places. Part of the data will remain at the signal cone centred around k=0 (i.e. the part with frequencies around ω=π(2n+1)/T) due to the first term of equation 17 and part of the data will be mapped to a signal cone centred around the Nyquist wavenumber k.sub.N (i.e. the part with frequencies around

(41) $\omega =\frac{2\pi n}{T}\hspace{1em})$
due to the second term in equation 17. FIG. 2 illustrates that, in comparison to conventional data (left part of FIG. 1), the data has been partially shifted to k.sub.N. Specifically, the data in the signal cone 1 centred around k=0 has not been shifted, but the data in signal cone 2 centred around k=k.sub.N has been shifted.

(42) Thus, from equation 12 it is clear that when polarity flips are used, substantially all the data from the source will be shifted. However, from equation 17 it is clear that when using time dither data will only be partially shifted.

(43) However, the inventors have realised that if one of the terms of equation 17 is known from recorded data, then the other term can be predicted using equation 17. This is a critical observation that makes time dither as useful as flipping polarities.

(44) Since it is not necessary to flip the polarity using time dither as the varying signature, time dither can be performed using conventional sources (such as airguns). Flipping polarity, on the other hand, may require the use of more specialist equipment, such as marine vibroseis.

(45) As can be appreciated, the theory behind the present method may be presented in numerous different ways. Another way of considering the origins of the effect of varying the source signature from shot to shot is that the recorded data then can be considered to consist of a sum of individual datasets, where each data set has one individual/specific source signature. Say, when the source signature is delayed in every second shot, the data can be considered to be the sum of two datasets: one without source delay, and the other with source delay. The full data will have a sampling frequency of k.sub.s. The two individual datasets then will have sampling frequency k.sub.s/2. This property leads to all the benefits in acquisition, processing, modeling and inversion that have been described in the invention.

(46) Example in Practice

(47) FIG. 3 shows an example data set from a seismic survey. Although the data comprise a common shot gather sampled at 6.25 m trace spacing, we will manipulate the data as if it were a common receiver gather where every second trace corresponds to a new source location (unrealistically densely sampled at 6.25 m source spacing). Only a small part of the data has been selected such that for instance all near offsets are missing. This will generate some noise artefacts when transforming the data into the fk space.

(48) FIG. 4 shows an fk plot of the data in FIG. 3. This is an fk plot of data gathered using a conventional shooting pattern (e.g. the left-hand side of FIG. 1). In this particular case, most of the data arrive with negative wavenumber. This is because the source is located in front of the spread. We can clearly see the outline of a signal cone bounded by the minimum observable apparent velocity of arrivals (water velocity). We see how some energy “bleeds” outside the signal cone. This is an artefact caused by the fact that we chose a small section of data. A more complete data set (such as split spread data set with near and far offsets) would be better focused within the signal cone. However, the data shown here are good enough to serve the purpose of illustrating our concept.

(49) FIG. 5 shows an fk plot where every second trace has opposite polarity to every second trace (e.g. +1, −1, +1, −1, +1, −1, etc.). As expected the signal cone has been shifted along the wavenumber axis to be centered around the Nyquist wave number. This is as shown schematically in the right-hand side of FIG. 1.

(50) FIG. 6, shows the fk plot of the same data but the polarity is flipped as follows: +1, +1, −1, −1, +1, +1, −1, −1, etc. It can be seen that the signal cone has been shifted to be centred around positive and negative half of the Nyquist wavenumber.

(51) FIGS. 7, 8, 9 and 10 show fk plots of the data after applying a time shift to every second trace of 10 ms, 20 ms, 40 ms and 200 ms respectively. For instance, the source may be a flip/flop source. Note how part of the data shift from being centred around wavenumber k=0 to the opposite end of the wavenumber axis, i.e., the Nyquist wavenumber. A notch pattern can be seen (in the following referred to as “ghosts”, although these have nothing to do with a sea surface ghost) where for certain frequencies all the data is shifted and for certain frequencies none of the data is shifted. This notching can be understood by looking at equation 14. For certain frequencies

(52) $(\hspace{1em}f=\frac{\left(2n+1\right)}{2T}$
where T is the dither) all the data will be shifted, and for other certain frequencies

(53) $\left(f=\frac{n}{T}\right)$
none of the data will be shifted.

(54) As discussed above, it is possible to remove these notches, and shift all the data so that is centred around the Nyquist wavenumber. This is shown in FIGS. 11 and 12 where equation 16 is applied to the data with a 10 ms time dither (FIG. 7) to illustrate how we can recover amplitude of a signal cone that has been shifted to the Nyquist wavenumber (even if the signal cone around k=0 is lost or completely masked in noise or other data). This estimate is also used to “reghost” the data (this terms as used here has nothing to do with the sea surface ghost problem) to fully remove all that is left at k=0.

(55) FIG. 11 shows the data of FIG. 7 where equation 17 has been used to properly “deghost” the signal cone centred around the Nyquist wavenumber so that the shifted data now again corresponds to the original data (but shifted to the Nyquist wavenumber).

(56) FIG. 12 shows the data shown in FIG. 11 where equation 17 has been used to properly “reghost” the signal cone centred around the Nyquist wavenumber and subtract it from the signal cone centred around k=0. We are left with a data set with one signal cone only that has been shifted to the Nyquist wavenumber.

(57) Thus, using time dither, the data is partially shifted. However, the non-shifted data can be shifted mathematically by understanding the theory behind the shifting.

(58) Now some applications of the present method are described, by way of example only.

(59) 1. Simultaneous Source Acquisition

(60) In one embodiment we have two source boats. The first boat shoots every second shot with opposite polarity. The other boat acquires data conventionally (i.e. with no varying signature). The recorded data, in a common receiver gather, will contain a superposition of the two data sets. However, after an fk transform, the data separates to opposite ends of the k axis in the fk spectrum (one cone centred at wavenumber k=0 from the conventional source and the other cone centred at +/− the Nyquist wavenumber from the varying-signature source). The two data sets can now be isolated and inverse transformed back to the space-time domain to obtain the data sets corresponding to each source boat separately. The data set where every second trace has opposite polarity can now be conditioned so that every trace has the same polarity.

(61) In another embodiment, one source is fired without a time shift whereas a second source is fired using a constant time dither (e.g., 10 ms as shown above) for every second shot. The data from the first source will always end up in a signal cone around k=0. However, the data from the second source will be split between two signal cones; one centred around k=0 and one centred around the Nyquist wavenumber in accordance with equation 14. The above theory shows how to:

(62) (1) Fully recover the data from the second source using the signal cone around Nyquist wavenumber only (through what resembles the “deghosting” operation discussed above).

(63) (2) Remove all energy from the second source that was left behind in the signal cone centred around k=0. In other words, the data from the first source is fully recovered.

(64) The concepts of these two embodiments can be generalized to more than two sources and to different varying signatures. For instance, by having a third source with a time dither on two consecutive shots and then no time dither on the next two consecutive shots, then time dithers on the following two consecutive shots, etc., we will obtain data with a new signal cone introduced, centred around half the Nyquist wavenumber.

(65) Note that even though large parts of the fk space are empty in conventionally acquired data, common receiver gathers typically are acquired sparse so that they alias already at frequencies inside the frequency band of interest. Using the technique described here, the two data sets will start to interfere at an even lower frequency because signal cones of the data from the various sources may overlap above a certain threshold frequency value. It is desirable to avoid this as much as possible. The inventors have found several ways to mitigate aliased and/or interfering data:

(66) a. Instead of separating the data in common receive gathers, data could be separated in another domain such as common offset gathers. Common offset gathers are largely flat and apparent velocities will be much higher compared to common receiver gathers and therefore separate much better after an fk transform, i.e. the signal cone will have steeper sides, and hence be narrower, and so will interfere less with other signal cones. As long as the sequence of modulating time shifts from trace to trace is maintained in such a gather, we will separate the data as desired in fk.

(67) b. Since the lowest frequencies in each signal data cone will not overlap with other data cones (due to the shape of the data cone), the lowest frequencies are always unaliased. Dealiasing the aliased higher frequencies can be carried out using known techniques. On such technique “Interpolation with priors” (Spitz, 1991; Ozbek et al., 2009; Vasallo et al., 2010; Ozbek et al., 2010) exploit the fact that (1) a model of an unaliased higher frequency can be predicted from the aliased data, (2) the use of a lower unaliased frequency to compute priors, and (3) an assumption such as that the data contains linear events only in fk. Such dealiasing will be very effective on the types of data that we propose to acquire also in cases of using a greater number of source boats than two.

(68) c. By removing the direct wave, waves guided in the water layer, water bottom refractions, etc. (e.g., by modelling), the width of the signal cone can be narrowed substantially so that the signal cones are better separated in fk and the method will be more effective.

(69) d. If data are recorded far away from a recording location, the signal cone on a common receiver gather will appear narrower as the azimuth range is limited. Finding appropriate gathers to sort simultaneous source data on can be used to ensure that at least one signal cone is narrower and separated better from the other(s).

(70) These mitigation methods are applicable to any application of the present method where multiple sources are used.

(71) 2. Seismic Interference Cancellation

(72) Seismic interference is the undesired influence of a different seismic survey conducted in the vicinity of the own seismic survey. Seismic interference (SI) is relatively easy to remove if the interfering seismic energy is arriving in the inline direction of the seismic survey. However, a particularly difficult case is seismic interference arriving from the broad side. Using the technique described in this report we can move the signal to be as far as possible in the fk spectrum from the SI.

(73) SI data often has a low frequency bias compared to the seismic data acquired. In order to remove as much SI as possible, when using time dither a large time shift should preferably be chosen similar to half the dominant period in the SI. We expect the SI application to work particularly well due to the band-limited nature of SI such that one can avoid interference with the data being acquired (low frequencies shifted away from the data along the wavenumber axis will fully fall outside the signal cone of the data acquired).

(74) Using the present method the operator can make sure that the recorded data set will always be acquired at opposite side of the k axis compared to the seismic interference after an fk transform independent of the arrival direction of the seismic interference. The interfering data will therefore be even easier to remove than the currently most benign case of inline interference. The appropriate signature (e.g. the polarity variation or deterministic time shift dithering) sequence can be chosen directly in the field when encountering seismic interference. For instance, if the SI is caused by another vessel shooting seismic waves, it may be possible to select the signature appropriately if the source trigger times of the other vessel are known.

(75) 3. Residual Shot Noise Attenuation

(76) Residual shot noise (RSN) is recorded energy that arrives from deep reflections, shear wave conversions, high order multiples or combinations thereof but that were generated from the previous shot. It is a principal form of shot generated noise that limits signal-to-noise in recorded data in cases where other noise types such as ambient noise are weaker. Therefore, in such scenarios, if we can reduce RSN we can either i) shoot seismic data quicker (leading to faster tow speed and therefore shorter records), ii) shoot more densely, or iii) we can always guarantee that the data will be of higher quality if we retain the same towing speed and shot density. The removal of RSN can therefore have a significant impact on the cost efficiency of a survey. Note that RSN is particularly problematic for low frequencies since low frequency data suffer less from attenuation in the Earth's subsurface and therefore require longer times to decay before we are ready to acquire a new uncontaminated shot.

(77) In one embodiment, the following method may be used to isolate residual shot noise when acquiring seismic data using one source boat. First, shoot two consecutive shots with the same polarity. Then shoot two consecutive shots with opposite polarity. Next again shoot two consecutive shots with the same polarity as the first two followed by two with opposite polarity, etc. After acquiring the data, multiply all shots with opposite polarity with −1 (or multiply all shots with positive polarity with −1) such that all traces now have the same polarity. Interestingly, the residual shot noise will have opposite polarity on every second trace. Because of this, after an fk transform, the residual shot noise ends up on the opposite side of the k axis compared to the desired signal and can be efficiently muted.

(78) In another embodiment, flip/flop sources may be used so that the time between consecutive flop shots is always the essentially same and essentially also the same as the time between consecutive flip shots. However, the time between a flip and a flop shot is different compared to the time between a flop and a flip shot.

(79) FIG. 13 illustrates how flip/flop data with these types of time shifts can be acquired with fully uniform shot positions. The top of FIG. 13 shows a conventional flip/flop source arrangement where the two stars represent the two airgun arrays that have the same inline offset but are shifted in the cross line direction. In the bottom of FIG. 13 additionally the flip source has been shifted compared to the flop source in the inline direction.

(80) As an example, consider a case where data are acquired with a tow speed of 2.5 m/s. In a conventional flip/flop acquisition data is shot every 10 s so that we obtain a distance between flop shots of 50 m and a distance between flip shots of 50 m as well. Flip and flop shots are perfectly staggered with respect to each other.

(81) In our method a slight time shift between flip and flop shots may be introduced as the time dither. For example, the time between flip and flop shots is 9.8 s and the time between flop and flip shots is 10.2 s. By staggering the sources in the inline direction as illustrated in the lower half of FIG. 10, it is possible to still acquire data on a fully uniform grid. All that we require is to stagger the sources by a distance that corresponds to the distance that the boat moves forward over 0.2 s which in our case is 0.5 m. Note that for the preferred staggering times of say 10 ms or 20 ms, this distance is so short that it can be ignored (2.5 cm in the case of a 10 ms time shift) such that we can continue to tow flip/flop sources as is conventionally done (top of FIG. 13).

(82) FIG. 14 illustrates a common receiver gather acquired using flip/flop shooting using a conventional technique (left) and the new method described here after backing off the time shift that was introduced during acquisition (right). The case illustrated shows where we tow the sources faster using the new technique such that the record length is shorter on the right hand of FIG. 14 compared to the left. Both signal and shot generated noise (RSN) from flop sources are coloured black whereas arrivals due to the flip source are coloured grey. In the conventional case we note that both signal and RSN are coherent and continuous from shot to shot. However, using the present method we note that whereas the signal becomes continuous from shot to shot, RSN suffers a time shift that is twice that of the original time shift introduced during acquisition. That is, if data were shot with 9.8 s between flip and flop shots and 10.2 s between flop and flip shots, RSN will be shifted by 0.4 s from trace to trace after backing of the original time shift such that the signal is continuous between shots. This effect can be exploited to move RSN away from signal centred around wavenumber k=0 to the opposite end of the wavenumber axis (Nyquist wavenumber) as described above. We can now fully remove the RSN without harming the signal after a suitable transform to the fk domain for instance. Note that the optimal choice of staggering times between flip and flop sources will depend on geology and the character of the RSN. It is likely that just as in the case of the SI application, we will benefit from focussing on low frequencies only (just as for SI, RSN tends to be particularly severe at low frequencies). Again, a particular advantage of the low-frequency bias is that we will be much less prone to problems with spatial aliasing.

(83) Whilst this has been discussed in terms of flip/flop sources, the same principle may be used for any source with a periodic varying signature.

(84) 4. Seismic Data Modelling and Reverse Time Migration (RTM)

(85) Seismic modelling engines such as finite differences (FD) form the basis of state-of-the-art modelling, imaging and inversion algorithms. Such modelling engines are extremely computational intensive and if so generating synthetic data using more than one shot point at a time could increase the efficiency significantly.

(86) It is clear that using the present method one can immediately recover unaliased synthetic data with two (or more) simultaneous sources using the techniques described herein. This is particularly the case if all but one of the sources only contain low frequencies up to the point where they would start to interfere with the other data, since in this case the generated data is always unaliased and can be recovered for sufficiently low frequencies.

(87) Thus, low-frequency data can be acquired at the same time as a conventional source and so—in terms of computing power—are effectively acquired for free. The low-frequency data is of low enough frequency so that that it will not interfere with any of the data from the other low frequency source(s) or the conventional source(s).

(88) Further, the above-discussed techniques relating to minimising interference and aliasing can be used to mitigate interference and aliasing issues between sources.

(89) 5. Broadband Seismic Acquisition

(90) In order to perform broadband acquisition, it has been proposed to use a dedicated low frequency source, such as a “sub-woofer” in combination with a conventional source (Berkhout (2012)). Using our invention we can acquire such “sub-woofer data” simultaneously with a conventional source that cover a little low frequencies but mostly intermediate and high frequencies. Acquiring the “sub-woofer data” flipping polarity at every second shot point can therefore be done without interfering with the conventional data at all (similarly to the modelling application described earlier). Alternatively, time dither could be used.

(91) Depending on the maximum frequency of the “sub-woofer data”, we can also choose to acquire it sparser without interfering with the conventionally shot data or without aliasing the “sub-woofer data” themselves. However, marine vibroseis are known to be inefficient at emitting low frequencies. Therefore, even if we have a purpose built low-frequency marine vibroseis we will likely benefit from shooting often to compensate for the weaker output.

(92) Thus, the multiple simultaneous sources can comprise at least one low frequency source and at least one conventional source.

(93) 6. Cost-Effective Acquisition of Shear-Wave Data

(94) Converted wave (shear) data can be acquired much more efficiently using the time dither, or polarity flipping, concept enabling record lengths that are similar to those of conventional pressure data. The procedure and benefits are analogous to those described under the RSN application outlined above.

(95) Note that both RSN and shear waves occur late in the record and in both cases the apparent wavenumbers are limited (waves mostly arrive close to the vertical) such that time dither will work particularly well.

(96) In the case of pressure and shear data acquisition we also benefit from the fact that shear data tend to be mostly arriving on the horizontal component in seabed recordings thus leading to more favourable signal-to-noise ratio in the separation process. Likewise pressure data dominate the pressure and the Z recordings.

(97) Finally, just as in the RSN application, we benefit from the fact that the recorded shear arrivals typically lack high frequencies and therefore are limited to lower apparent wavenumbers, and so are less likely to interfere.

(98) 7. Deghosting and Source-Side Gradients for Interpolation

(99) By applying the dithering sequences to different sub-arrays within an airgun array, it is possible to separate the responses from sub-arrays such that horizontal gradients on the source side can be computed. These are useful for source-side deghosting and other applications.

(100) If the simultaneous source concept is used for sources (or sub-arrays) that are closely located to each other, one can estimate spatial derivatives in the vertical and horizontal directions. Note that we can use different signature sequences to have three (or even more sub-arrays) firing at the same time with different dithers that then can be separated. From these data spatial derivatives of the wavefield on the source side can be computed for a range of applications, for instance for: vertical derivative can be used for source-side deghosting and/or horizontal derivatives can be used for spatial reconstruction of the wavefield on the source side (Robertsson et al., 2008).

(101) Essentially, in this case, the array (or sub array) is treated as a comprising multiple sources. If the signature of each source or each sub array is varied in accordance with the present method, it is possible to know what recorded data came from each source (or sub array). Knowing this can greatly ease deghosting and source-side gradient calculations.

(102) It should be apparent that the foregoing relates only to the preferred embodiments of the present application and the resultant patent. Numerous changes and modification may be made herein by one of ordinary skill in the art without departing from the general spirit and scope of the invention as defined by the following claims and the equivalents thereof.

REFERENCES

(103) Berkhout, A. J. (2012). Blended acquisition with dispersed source arrays. Geophysics, 77(4), A19-A23. Özbek, A., Özdemir, A. K., & Vassallo, M. (2009, January). Interpolation by matching pursuit. In 2009 SEG Annual Meeting. Society of Exploration Geophysicists. Özbek, A., Vassallo, M., Özdemir, K., van Manen, D. J., & Eggenberger, K. (2010). Crossline wavefield reconstruction from multicomponent streamer data: Part 2—Joint interpolation and 3D up/down separation by generalized matching pursuit. Geophysics, 75(6), WB69-WB85. Robertsson, J. O. A., I. Moore, M. Vassallo, A. K. Özdemir, D. J. van Manen and A. Özbek, 2008, On the use of multicomponent streamer recordings for reconstruction of pressure wavefields in the crossline direction: Geophysics, 73, A45-A49. Spitz, S. (1991). Seismic trace interpolation in the FX domain. Geophysics, 56(6), 785-794. Vassallo, M., Özbek, A., Özdemir, K., & Eggenberger, K. (2010). Crossline wavefield reconstruction from multicomponent streamer data: Part 1—Multichannel interpolation by matching pursuit (MIMAP) using pressure and its crossline gradient. Geophysics, 75(6), WB53-WB67. Yilmaz (2001): Seismic Data Analysis: Processing, Inversion, and Interpretation of Seismic Data, Investigations in Geophysics: SEG