PetroleumSeismology.com
Home » Images » Chapter 10

Images in: Chapter 10

1 of 5 Next»

Free-surface multiples

Images - Chapter 10 An illustration of the process of attenuation of free-surface multiples and de-ghosting as a transformation from (a) a physical experiment to (b) a hypothetical experiment.

Our objective in multiple attenuation (also known as demultiple) is to transform the data recorded over a medium, which consists of a solid half-space overlain by a water layer with an air-water free-surface interface, to hypothetical data corresponding to the same model without the free surface; i.e., the water layer is now infinite, as illustrated in here. This transformation corresponds to removing ghosts and free-surface multiples from the data.

A marine seismic line acquired by Total

Images - Chapter 10 Solid line: Source signature with its amplitude and phase spectra derived by using the portion of data above the second-order multiple of the sea-floor reflection (S111). Dashed line: Source signature with its amplitude and phase spectra derived by using the entire data. This source estimation corresponds to shot 100.

The data here are a marine seismic line acquired by Total (now TotalFina). As with most present real towed-streamer data, the data are limited to only a single pressure-field measurement. We have numerically predict the vertical component of the particle velocity needed for the application of the Kirchhoff series. We have also assumed in our application of this series that pressure does not contain ghosts by treating any ghost effects as part of the source signature. The direct wave was also muted.

This figure shows a source wavelet with its amplitude and phase spectrums by using the portion of data above the second-order multiple of the sea-floor reflection (S111) and by using the entire data. This source estimation corresponds to shot 100. Notice again that the limitation to the portion above S111 has essentially corrected only the amplitude of the wavelet. As in the synthetic example, the phase of the wavelet is largely unaffected. For the remainder of our discussion in this section, we will use the source signature estimated, based on the limited portion of the data above S111.

Shot gathers after free-surface multiple elimination

Images - Chapter 10 (a) Shot gathers after free-surface multiple elimination. (b) Shot gathers before free-surface multiple elimination.

The nearest recorded offset in this dataset is 200 meters. However, the missing near offsets (0 to 200 meters) are required for the inverse scattering demultiple. To fill this offset gap, we have used the extrapolation technique proposed by Vershcuur. It consists of first applying an NMO correction in order to align events horizontally in time and then filling the offset gap by fitting amplitude curves across constant times.

This figure shows shot gathers before and after multiple suppression. The first five terms of the Kirchhoff series were used for the multiple removal. Multiple energy is reduced significantly, and multiples are removed from the interfering event at 1.5 seconds.

Stacked seismic section

Images - Chapter 10 Stacked seismic section before free-surface multiple removal.

Stacked seismic section

Images - Chapter 10 Stacked seismic section after free-surface multiple removal.

In the stacked sections in the previous figure and here, one can notice several locations where multiples interfere with primary events. We have indicated three examples of primary-multiple interference by arrows. The first example (arrow A) is a multiple which lies at 1.83 seconds on CMP 800 and rises to 1.72 seconds at CMP 100. At CMP 580 a slight discontinuity in the event suggests that in this region the multiple overlies a primary event. After the demultiple, the multiple event is attenuated along the length of the seismic section, where there is no primary interference, but leaves a primary event between CMP 300 and CMP 700, which form the top of a mound structure. The second example is illustrated by arrow B. Before the demultiple, we can observe a double set of strong free-surface multiples dipping from right to left, starting at 2.2 seconds on CMP 100. After the demultiple, the free-surface multiples are well attenuated, with a primary event revealed below the lower free-surface multiple between CMP 200 and CMP 350. Arrow C marks the last example of the structure emerging after the demultiple, where a small mound structure at 1.15 seconds emerges from the demultiple seismic section. The interpretation of these three events was supported by other independent analyses. Also, in the case of the example marked by arrow B, we would expect that the estimated source wavelet would allow either the attenuation of both free-surface multiples or fail for both. Hence we concluded that the event between CMP 200 and CMP 350 on the demultipled seismic section is a primary.

The above examples show that the demultiple process has suppressed multiple reflected energy if it is present but has not significantly attenuated the primary energy along with it.

Stacked section

Images - Chapter 10 Stacked section before and after free-surface multiple attenuation.

The Troll dataset considered here is also a 2D line. It was acquired in very bad weather in 1994 by Western-Geco, which was then Geco-Prakla. So this dataset contains a significant swell noise due to bad weather. We have applied a low-cut filter up to 12 Hz to reduce this noise to an acceptable level. The other preprocessing steps were the mute of the direct wave and the √{t} amplitude scaling for the 3D-to-2D amplitude correction.

The Kirchhoff series requires that the input data contain near offsets, including the zero-offset. In the Troll dataset, the nearest offset is 37.5 m, with 18.75 m spacing between offsets. To fill up the two missing offsets (0 m and 18.75 m), we decided to duplicate the nearest offset, thus avoiding extrapolation for the missing near offsets.

As in the Barent Sea example, we use the Kirchhoff series and numerically predict the vertical component of the particle velocity. We have also assumed in our application of this series that pressure does not contain ghosts by treating any ghost effects as part of the source signature. The direct wave was also muted.

To analyze the results of the Kirchhoff series on the Troll data, we performed an NMO stack of data before and after the demultiple. This figure shows NMO-stacked sections before and after the demultiple. The only processing difference between the two seismic sections is the Kirchhoff demultiple process. We have used the first five terms of the Kirchhoff series.

The Troll area is fairly horizontally flat. So, with a careful velocity analysis we expect the NMO stack to reduce a significant amount of multiple energy by the differential moveout. Yet by comparing the sections before and after the demultiple, we can still see a significant improvement after the application of the Kirchhoff series.

We have highlighted three examples of primary-multiple interferences in the next two figures.

Multiple/primary interference

Images - Chapter 10 Example of multiple/primary interference.

The undulating primaries around 0.95 seconds are distorted by the multiple energy. After the demultiple, the multiple events are well attenuated, and the undulating primaries are now clear. We have highlighted a second example of primary-multiple interferences around 1.05 seconds in the same figure. Using raw data, we can observe a free-surface multiple right through the section. After the application of the Kirchhoff series, the free-surface multiple is well attenuated, and an incoherent primary is revealed

Multiple/primary interference

Images - Chapter 10 Example of multiple/primary interference.

This figure shows another example of primary-multiple interferences. This time, we are inside the Troll gas reservoir. We have highlighted a dipping event inside the reservoir. Notice that this event is totally obscured by multiples in the raw data.

NMO-corrected CMP before and after free-surface multiple elimination

Images - Chapter 10 NMO-corrected CMP gather 600 before and after free-surface multiple elimination.

For modern seismic processing tools (like prestack waveform inversion, AVO and AV0-A analysis, etc.), it is rather important to analyze the effectiveness of multiple attenuation on prestack data, especially on CMP gathers. This figure and the next figure show two NMO-corrected CMPs before and after the demultiple. Again, we have highlighted some examples of primary and multiple interferences. At 1.5 seconds in raw data, the primary is completely obscured by various multiple interferences. After the demultiple, the primary is much more visible. Another example is highlighted at 2.05 seconds. The AVO of this primary is distorted by free-surface multiple interferences, especially at large offsets. After the demultiple, AVO behavior is much more clearly defined. The predicted free-surface multiples for these two CMPs are shown in latter.

Notice also the ringings on raw data, especially at near offsets. These ringings interfere with the primary and have the same NMO as primaries at near offsets. As we can see in this figure and in the next, most of this ringing energy is well attenuated by the Kirchhoff series, whereas other multiple algorithms might require an inner mute of these ringings.

There is a strong event highlighted in this figure and in the next at about 3.0 seconds which was originally interpretated as a free-surface multiple. Based on the predicted multiple wavefield, shown just after the demultiple figures, it is easy to see that this event is not predicted as a free-surface multiple; therefore it is either a primary or an internal multiple.

NMO-corrected CMP before and after free-surface multiple elimination

Images - Chapter 10 NMO-corrected CMP gather 700 before and after free-surface multiple elimination.

1 of 5 Next»