Advances in deepwater data acquisition improve subsalt imaging.

Today, some of the highest-profile targets in the Gulf of Mexico are in deepwater basins beneath complex salt masses. As such, exploration and production risks are huge. Seismic data should play a key role in mitigating those risks, but unfortunately the quality of the data is often very poor - even in state-of-the-art surveys.
Much attention over the last few years has concentrated on data processing, with special emphasis on multiple suppression and, of course, depth imaging. Only within the last couple of years has attention been extended to improving the data acquisition product.
The quest for resolution
A popular thrust in the seismic method has been the pursuit of resolution, which requires high frequencies. In towed streamer surveys, a key phenomenon influencing the recordable frequency range is ghosting.
Ghosting is caused by interference of the upgoing wavefield with a polarity-reversed copy of itself that is sent back in the downgoing direction after scattering off the air/water interface. The amplitude spectrum of the ghost filter possesses notches at zero Hz and at regular intervals along the frequency axis. The locations of those notches depend on the depths of the source and streamers - the shallower the source and streamers, the higher the frequencies where the notches occur.
The first non-zero notch is typically considered to be the upper limit of useful bandwidth. Forcing that first non-zero notch to occur as high on the frequency axis as possible is often a key objective in survey design. With this in mind, the desire is to tow the source and streamers at shallow depths, about 16 ft to 20 ft (5 m to 6 m). However, there is a sobering issue associated with this. If the sea state is such that swell noise is present, the severity of that recorded noise worsens with shallower depths of streamer tow.
Therefore, the best way to keep the signal-to-swell-noise ratio at an acceptably good level is to tow the streamers a little deeper than would otherwise be desired. Typically, this means that streamers are towed at depths of 30 ft to 36 ft (9 m to 11 m). The pass bands for ghost filters associated with the desired tow depth of 16 ft and the pragmatic tow depth of 30 ft are compared in Figure 1. Using the 10-dB down points as reference, the resolution in the 30-ft case is only about half as good as that in the 16-ft case.
The high-resolution approach
Recently, single-sensor sampling has enabled the swell noise problem to be addressed more effectively.
Swell noise actually propagates as a coherent train across the seismic record. However, propagation velocity is extremely slow, meaning that in conventional records, where the group center trace interval is 41 ft (12.5 m) or so, the noise train is too aliased to be suppressed by velocity filters. However, with single-sensor technology, the trace interval is 10.2 ft (3.1 m). This interval is sufficiently dense to allow adaptive velocity filters to be effective in data processing.
The implication is that single-sensor technology can tolerate the recording of more swell noise, thereby enabling the streamers to be towed at a shallower depth, resulting in better resolution. This high-resolution implementation of technology has been the key strategy for these surveys.
Despite these recognized successes, the customer community did not immediately acknowledge this methodology to be a viable approach for use in subsalt imaging in the Gulf of Mexico.
The need for low frequencies
The reason for this was that seismic surveys using various state-of-the-art systems had already demonstrated it was extremely difficult to produce any kind of credible image in complex subsalt regions. Whenever success was achieved, it was only at the low end of the frequency spectrum. High-resolution acquisition was considered to be naive and inappropriate for the complicated subsalt environment.
To address that concern, frequency-dependent issues in complex imaging were investigated using numerical experiments. Aside from the inelastic attenuation associated with wave propagation, it was determined that a substantial amount of high-frequency signal is lost in imaging when migration velocity models are too smooth or inaccurate.
Therefore, a pragmatic lesson learned was that since only the low frequencies could survive the imaging process (or least, the first few iterations of that process), it was imperative that those frequencies be acquired. That is, acquisition operations needed to generate, preserve and record the low-frequency data.
To that end, a series of experiments was conducted over four fields in the Gulf of Mexico from December 2003 through April 2004. Some were partially funded by client oil companies, and all were proprietary at that time. Two key acquisition strategies were tried. Those strategies are described below.
Noise
Aside from the previously discussed swell noise, two other types of noise can also plague the low end of the spectrum. These arise from the strumming action of the streamer stress member and from bulge waves that travel in the streamer skin. All three types of noise are coherent and can be addressed by the adaptive velocity filtering empowered by single-sensor sampling. However, there are two other aspects common to all three noise components that can be exploited further in the data acquisition strategies. First, they all are source-independent. That is, they are created by rough sea conditions, not the seismic source. And second, they are all ghost-independent. That is, they are not part of the upgoing/downgoing wavefield complex, so they are not filtered by the ghost action.
Acquisition strategy #1
The fundamental principle in strategy #1 relied on the fact that increasing the low-frequency source strength would raise the signal level without increasing the noise. (The noise would not be raised because it is source-independent.) The low-frequency source strength was increased in two ways. First, the total volume of the airgun array was simply bumped up to the maximum that could be achieved with the available hardware. Second, the firing times of the different-sized airguns comprising the array were adjusted so that the first bubble oscillations were synchronized. This served to funnel more of the available source energy into the low frequencies. Because a bubble-tuned wavelet is not minimum phase, it was important to record the pulse at each shot using the calibrated marine source feature of the acquisition system. This capability enabled proper designature to be performed later in the data processing sequence.
In Figure 2, results from one of the experiments are compared with reprocessed legacy data. The yellow shading in the figure denotes salt. Encased within the salt mass is a prolific oil field. The reservoir layers trend from the lower left to the upper right. These can be seen very clearly in the low-frequency test, whereas they are masked by interfering noise in the conventional result.
Because this strategy can bring low-frequency imaging benefits to areas where previous seismic surveys were worthless, it clearly has a place in an acquisition portfolio. However, the obvious potential disadvantage of the strategy is that only low frequencies are imaged, leaving no option to extend the bandwidth to higher frequencies after the low-frequency problem is solved. That drawback was addressed by strategy #2.
Acquisition strategy #2
The fundamental principle in this strategy relied on the fact that removing the streamer ghost would pass more low-frequency signal without increasing the noise (The noise would not be increased because it was not filtered by the ghost in the first place).
This was done by measuring the pressure field at two different depths. This enabled separating the signal-rich upgoing wave from the downgoing wave. Discarding the downgoing wave removed the ghost notches and thereby allowed retrieving the signal at all frequencies. The pressure field was measured at two different depths by towing the streamers in pairs, with one streamer over the other. The streamer-steering feature of this system made this possible.
The advantage of recording the entire spectrum in complex subsalt regions is that once the very low frequencies lead the way to initial imaging solutions, higher and higher frequencies can gradually be woven into the result.
Conclusions
The road to better imaging should start with improved data acquisition, and that particular attention should be paid to cherishing the low frequencies comprising the signal. This philosophy is slightly contrary to the high-resolution emphasis given in the majority of single-sensor surveys conducted previously in many areas of the world.