Multiple Attenuation
Spectrum is able to offer its clients a wide range of demultiple solutions to help resolve the various random and coherent noise issues presented. Below are some of the multiple attenuation techniques available:
Parabolic Radon Multiple Attenuation
After normal moveout (NMO) correction, a shot or CMP gather can be thought of as being composed of many different parabolic or hyperbolic events. Each curve can be described by the zero-offset time and the residual moveout time at a certain offset (usually the far offset). Radon decomposes an ensemble into its many parabolic or hyperbolic components in the Radon transformed domain, mutes out unwanted parts of the transformed domain according to the user specification and inverts the muted transform back into the time-offset domain.
Multiple removal can be performed in one of two ways:-
(a) Model the multiples - i.e. mute the primaries in the transformed domain, invert transform back and subtract the multiples from the input data. This will only suppress the multiples, the random noise still remains. Therefore, the output ensemble appears more like the original.
(b) Model the primaries - i.e. mute the multiples (noise) in the transformed domain. This will lead to the suppression of both multiples and random noise.
“High Resolution Radon” modules are available which work in the Fourier domain. These programs use weights to focus the decomposition onto its most significant spectral components. The weights are obtained by using the radon spectrum obtained from the previous frequency. A scalar is used to introduce a viscosity, which prevents undesired variations caused by random noise.
Our processing software performs high resolution multiple removal in a single pass, both modelling and subtracting the multiples. Two distinct approaches are available for solving the least squares inversion problem. Firstly the Fourier domain method described above and secondly by using the Non-uniform Discrete Fourier Transform approach (NDFT). In the NDFT approach the number of ray parameters is frequency dependent and determined internally by the module.
SRME
Spectrum is a member of the Delft consortium, our Surface Related Multiple Elimination (SRME) program is based on the published work of Verschuur & Berkhout (1997).
In the SRME process the surface related multiples are estimated by convolving each trace on each input shot with the shot at the trace location and stacking in a common receiver sense. This yields the 'shot record' of the next Taylor term, which can be used in subsequent acoustic model building or adaptive subtraction. This multiple model is removed from the original data using Optimum Least-Squares (OLS) filtering over whole ensembles. A Wiener-Levinson filter adjusts the amplitude and phase of the estimated multiples to match the original data. The filter which best subtracts all multiples is derived for each ensemble in the SHOT or RECVR domain. The matched multiple estimate is then subtracted from the original data.
Resources: Verschuur, D.J., and Berkhout, A. J., 1997, Estimation of multiple scattering by iterative inversion, Part II: Practical aspects and examples, Geophysics, 62, 1596-1611
Tau-p Decon
Multiples are only truly periodic in the X-T domain at zero offset, so pre-stack deconvolution is of limited use as a multiple attenuator. By transforming data into the linear Tau-p domain multiples can be made periodic for all values of P and effectively attenuated with predictive deconvolution. This technique is particularly effective in shallow water areas where muting in the tau-p domain can assist with linear noise attenuation.
MULLOCK (Targeted Multiple Attenuation)
MULLOCK is useful in instances where there is little to no velocity separation between the multiple and primary trends in which case traditional methods such as Radon are ineffective. It also works well in situations where the aliasing of multiples occurs. The process can be parameterized to be spatially variant so that when the dip of the primaries and multiples coincide, it can be effectively turned off.
MULLOCK attenuates peg leg multiples on stacks or common offset planes. The target multiple is interpreted and flattened before being decomposed to eigenimages. Then a limited range of eigenimages are selected for reconstruction before the flattening is reversed. The main assumption is that after flattening the multiples are represented by the dominant eigenimages, which can be omitted from the reconstruction.
FK Multiple Attenuation
A multiple NMO trend is picked or a percentage of the primary NMO (e.g. 95%) is taken and applied to CDP gathers. The data is then transformed to the FK domain and a filter, designed to remove an area of the +K quadrant of FK space, is applied prior to inverse transform back to XT space and removal of multiple NMO.
