Separated wavefield imaging of surface multiple energy Andrew Long, Petroleum Geo-Services, 23 May 2013

A new seismic imaging methodology introduced here exploits the illumination corresponding to surface multiple energy, and thus exploits what has historically been treated by the seismic industry as unwanted noise. Whereas a strong cross-line acquisition footprint affects any shallow 3D data using conventional processing and imaging, the new results can yield spectacular continuous high resolution seismic images, even up to, and including the water bottom.

Multiples and illumination

Figure 1 is a schematic illustration of various primary reflection, surface ghost, and both surface and internal multiple ray paths for towed streamer seismic data. In the historically ideal scenario, primary reflection data is recovered that includes no surface ghost effects and no multiple reflections from the earth. Terminology used below describes the 'up-going' pressure wavefield as that scattered up from the earth and yet to encounter the free-surface of the ocean. The 'down-going' pressure wavefield is the time-delayed version of the up-going wavefield that is reflected downwards from the free-surface of the ocean with opposite polarity (also referred to as the 'receiver ghost' version). Conventional hydrophone-only streamers record a continuously interfering combination of the up-going and down-going pressure wavefields – the total pressure wavefield.

In the approach described here, one-way wave equation pre-stack depth migration (WEM) is reconfigured to use the up-going and down-going wavefields to image the earth with surface multiple data that would historically be treated as unwanted noise (refer also to Figure 2). The up-going and down-going wavefields are derived from wavefield separation of dual-sensor streamer data. Surface multiples provide laterally more extensive illumination of the earth than primary reflections for a conventional 3D towed streamer geometry, particularly for shallower geology.

Figure 3 shows 3D ray tracing-based modelling of the illumination at a target surface roughly 500 m below the water bottom in a 3D model interpreted from real seismic data. In each case, 10 consecutive shots were modelled for three adjacent sail-lines. The dual-sensor streamer spread was 10 x 6,000 m streamers at 100 m separation and with dual-source shooting. First order surface multiples from the target interface remove the classic far offset illumination gaps modelled using primary reflections. Higher order multiples will illuminate the area between adjacent sail-lines with even higher density. The imaging process described here (Separated wavefield imaging: SWIM) exploits exactly this illumination, but provided by all orders of surface multiples – not only the first-order multiples. In principle, the cross-line illumination extent for surface multiples can be almost as large as the streamer spread width itself (the cross-line receiver distribution), in contrast to the CMP-based illumination coverage of primaries which is generally about half the width of the streamer spread.

Hence, some of the major contributors to the shallow cross-line acquisition footprint (loss of surface fold and target illumination coverage in particular) are mitigated with the 3D imaging solution (SWIM) described here. Another attractive aspect for the near-term development of this solution is that the cross-line acquisition footprint generally affects only shallow data (0-1 seconds at most, even for ultra wide-tow spreads), so the target depth range of interest is unlikely to be affected by cross-talk imaging artifacts associated with very high order surface multiples.

SWIM method and results

The forward propagated source wavefield in one-way shot profile WEM was replaced with the down-going wavefield derived from wavefield separation, and the back propagated receiver wavefield was replaced with the up-going wavefield. A deconvolution imaging condition can be used to construct the subsurface image whilst suppressing artifacts.

All processing and imaging applied here followed a simple workflow. Wavefield separation was applied in the shot domain, followed by simple noise attenuation, shot profile separated wavefield WEM (SWIM), and then fast-tracked to stack. As an initial test, five adjacent sail-lines from a 3D dual-sensor towed streamer dataset in the Browse Basin, Australia, were imaged over a time window of 0-3 seconds and up to a maximum frequency of 30 Hz. The dual-sensor streamer spread was 10 x 8,100 m streamers at 100 m separation, 15 m streamer depth and with dual-source shooting. Note how the cross-line acquisition footprint has been mitigated in Figure 4 when the surface multiple illumination is imaged.

Following these preliminary test results, a full-scale test for separated wavefield imaging was pursued with a 400 km² extract from a 3D dual-sensor towed streamer survey over the Tenggol Arch area in offshore Peninsula Malaysia. Water depth is approximately 70 m. This survey was acquired in 2011 using a 12 x 4,050 m dual-sensor streamer spread with 75 m separation, 15 m streamer depth and with dual-source shooting. Nominal streamer spread width was therefore 825 m. Imaging was pursued up to 60 Hz, ramping off to a maximum of 80 Hz. This choice of frequency range was made simply because of the experimental nature of the test, and a much higher frequency range could be selected instead. A final Kirchhoff pre-stack time migration (PSTM) volume was already available, so the depth imaged results were stretched back to two-way time (TWT) for comparison (refer to Figures 5 and 6).


All the results presented here are very encouraging. Considerable flexibility exists within the SWIM methodology being used. The imaging operator can be adjusted in terms of numerical complexity according to the lateral variability of the velocity model, full anisotropy can be accounted for, and angle gathers can be created to assist with velocity model building. Thus, a simple V(z) velocity model can be used for a robust first-pass SWIM imaging effort applied to raw field gathers after wavefield separation. Subsequent iterations would benefit from velocity model building in the image domain. The only assumption is that true wavefield separation of dual-sensor streamer data has been completed in pre-processing. Although not shown here, a shallow window from the depth imaged results were stretched back to time and successfully merged with time domain Kirchhoff migrated results from conventional imaging. The resultant 3D data cube therefore was achieved cost effectively and with negligible acquisition footprint effects all the way up to, and including, the seafloor reflection event.

Although the maximum frequency imaged in the Australian and Malaysian examples presented here was 60 Hz, there is no theoretical upper limit. High-order multiples introduce increasing levels of cross-talk noise, but this can be ignored for the depth range affected by the cross-line acquisition footprint – even for shallow water. The paradigm shift is that the opportunity exists to improve 3D marine seismic survey efficiency whilst the very shallow seismic images are improved in terms of both vertical and lateral resolution!


PETRONAS Carigali Sdn. Bhd. and Lundin Malaysia are thanked for support in this project. PGS is thanked for permission to publish this methodology and the results. 

comments powered by Disqus